{"id":"6fa302e1-1989-4d7e-b388-407cf2e8d6e3","arxiv_id":"1908.03307","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For a dense family of analytic planar domains, Steklov eigenfunctions have fixed-size zero-free balls in the interior, so their nodal sets are not dense at any shrinking scale.","lead":"Steklov eigenfunctions of a dense family of analytic 2D domains have nodal sets that stay away from certain fixed-size balls, so the nodal sets are not dense at the tiny wavelength scale. This answers an open problem in spectral geometry and shows interior Steklov eigenfunctions can behave very differently from Laplace eigenfunctions.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Lemma 4.1 depends on an imported [GT19, Cor. 1.3] exponential-concentration estimate that is not derived in this paper; this is the least secure link in the chain from tunneling to Theorem 1.","rationale":"The reader's weakest-assumption analysis points to the same place. My pass through the rest of the argument found no independent obstruction: the BBLCN approximation (Lemma 2.1 and Corollary 2.2) is plausible from polynomial approximation and the implicit function theorem; Theorem 3's recurrence (3.2) genuinely gives the superpolynomial off-diagonal decay in Lemma 3.1; and Theorem 2's Taylor-coefficient estimates are routine once Lemma 4.1 supplies the low-frequency mass bound. The only step where the proof rests on an externally supplied result with exactly the right quantitative form is Lemma 4.1's use of [GT19, Cor. 1.3]. This does not change the verdict because the cited result is published and the use is consistent with known Steklov coefficient concentration, but it keeps confidence at moderate until the corollary is checked. The m=0 application and the terse IFT details in Corollary 2.2 are repairable and do not threaten the central claim.","tokens_in":18190,"tokens_out":24791,"duration_ms":264702,"concrete_test":"Independently check [GT19, Cor. 1.3] against the use made in Lemma 4.1: obtain the published statement and verify that it yields, for σ>3m and for every Steklov eigenfunction of a bounded simply-connected analytic domain, ∑_{|k-m|≤2σ}|hat u(k)|^2 ≥ (1−C e^{−σ/C})||hat u||_2^2 with hat u defined by (1.5). If the corollary gives a different window or an upper bound instead, recompute Lemma 4.1 for fixed m with σ large and determine whether A_m^2 ≥ c e^{-Cσ}||hat u||_2^2 still follows; if it does not, Theorem 2's estimate (1.8) loses its uniform denominator and Theorem 1 would not be established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Lemma 4.1 (Section 4) is the bridge from the tunneling condition to Theorem 2. Its proof begins with the imported estimate [GT19, Cor. 1.3]: for σ>3m, ∑_{|k-m|≤2σ}|hat u(k)|^2 ≥ ||hat u||_2^2 (1−C e^{−σ/C}). Combined with Lemma 3.1, this gives the lower bound e^{-Cσ}||hat u||_2 ≤ A_m, which Theorem 2 then uses to control the high-frequency tail (via m=0) and to bound the denominator of the approximation error. The tail control and the fixed-radius positivity in Theorem 1 collapse if this estimate is absent or weaker than stated. The estimate is external to this manuscript, and it is co-authored by the second author, so a reader cannot certify it from the text alone. In particular, the proof needs the m-centered window and the m=0 case; if the published corollary states only concentration near |k|≈σ or gives an upper rather than a lower bound on low-frequency mass, Lemma 4.1 would not follow as written. This is a genuine load-bearing import, not an internal contradiction; I have not found an independent flaw in the BBLCN construction or in The proofs of Theorems 2 and 3.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies Steklov eigenfunctions in bounded simply-connected planar domains with analytic boundaries. It constructs a dense family of such domains for which, for every sufficiently large Steklov eigenvalue, every eigenfunction has a ball of sigma-independent radius on which it does not vanish, thereby giving a negative answer to Open Problem 10(i) of Girouard and Polterovich (J. Spectr. Theory 2017). The proof combines an approximation theorem by boundary-band-limited conformal domains (Corollary 2.2), a new 'tunneling' condition (Definition 1.1) shown to hold for BBLCN domains (Theorem 3), an exponential lower bound on low-frequency Fourier mass (Lemma 4.1), and a finite-mode approximation of interior eigenfunctions (Theorem 2). The paper also contains numerical experiments for elliptical and kite-shaped domains.","tokens_in":14,"tokens_out":16452,"duration_ms":345565,"significance":"If the main theorem holds, the result is significant: it establishes that Steklov eigenfunctions on a dense class of analytic planar domains can have slowly oscillating, sign-definite regions of fixed size, in stark contrast to high-energy Laplace eigenfunctions. The tunneling mechanism is a new and interesting idea, and the construction of BBLCN approximating domains is explicit and uses no fitted parameters. The proof of Theorem 3 is a clean recurrence argument. However, the central chain depends on an imported spectral concentration estimate from [GT19] and on a quantifier in the statement of Theorem 1 that is stronger than what the proof appears to establish; both points need to be addressed before the paper can be accepted.","major_comments":[{"comment":"The statement 'there exists a point x_sigma in B(x0,r0) such that ... each Steklov eigenfunction phi_sigma of eigenvalue sigma satisfies |phi_sigma|>0 on B(x_sigma,r1)' requires a single x_sigma common to all eigenfunctions in the eigenspace of sigma. The proof, around equations (4.4)-(4.10), chooses x0 from the maximum of the finite-mode approximation tilde u_{sigma,delta,m} for one eigenfunction, so it proves the conclusion only for that eigenfunction. If the Steklov spectrum of the constructed domain has an eigenvalue of multiplicity at least two, no common point can work: given two independent eigenfunctions positive at a proposed x_sigma, a suitable linear combination vanishes at x_sigma. Thus the theorem as stated is either false for eigenspaces of dimension greater than one or at least not proved. Please restate the theorem in the per-eigenfunction form used in the abstract, or add a genericity argument ensuring the constructed domains have simple Steklov spectrum.","section":"Section 1, Theorem 1"},{"comment":"The proof of Lemma 4.1 imports the estimate [GT19, Corollary 1.3] in the form sum_{|k-m|<=2sigma}|hat u(k)|^2 >= ||hat u||_{ell^2}^2(1 - C e^{-sigma/C}) without stating its precise hypotheses or proof. This estimate is load-bearing: it is the only source of the exponential lower bound e^{-C sigma}||hat u|| <= A_m, and without it the high-frequency tail estimate in Theorem 2 and the fixed-radius positivity in Theorem 1 fail. Since the corollary is external and is co-authored by the second author, a reader cannot certify it from the text alone. Please state the exact statement, confirm that it applies to the mapped Steklov problem on the unit disk with the weight |partial_z f|, and verify the m=0 case, which is used in the proof of Theorem 2 even though Lemma 4.1 as stated is for m>0.","section":"Section 4, Lemma 4.1"},{"comment":"The displayed application of Lemma 3.1 appears to have an indexing error: the sums are written with factors C_0^{2k} and C_0^{2|k|}, whereas Lemma 3.1 gives the bound C_0^{2|k-m|} for the coefficients in the window |k-m|<=2sigma. With the printed exponents, the subsequent bound by (2(2C_0^{4sigma+2}-1)/(C_0^2-1)) A_m^2 does not follow, particularly when m is large. This is likely a typographical error, but it must be corrected because the lower bound for A_m depends on this step.","section":"Section 4, proof of Lemma 4.1"},{"comment":"The proof of Theorem 1 uses the C^1 bound of tilde u_{sigma,delta,m} on a ball B(x0,r_{m,delta}) but the preceding estimates were obtained on B(0,delta). To ensure B(x0,r_{m,delta}) is contained in the unit disk and that the image f(B(x0,r_{m,delta})) lies in Omega_1 with the stated ball inclusion B(x_sigma,r1) subset B(x0,r0), the proof should specify the choice of delta and r_{m,delta} more carefully, e.g. taking 2delta<1 and r_{m,delta}<delta. The current text says r_{m,delta}<delta but does not state the required containment in the domain of definition of the approximation.","section":"Section 4, proof of Theorem 2 and Section 1, Theorem 1"}],"minor_comments":[{"comment":"The tunneling condition does not explicitly quantify over eigenfunctions; please state that the inequality is required for every Steklov eigenfunction u_sigma (or for every member of an orthonormal basis).","section":"Definition 1.1"},{"comment":"The notation switches between u_{sigma_j}, u_sigma, and tilde u_{sigma,delta}; please define these consistently near (1.3), (1.7), and Theorem 2.","section":"Notation"},{"comment":"The proof approximates w by a polynomial p_epsilon and then factors p_epsilon = beta_0 prod (z-beta_i)^{N_i}. To ensure the zeros satisfy |beta_i|>1 rather than merely |beta_i|>=1, one should note that a small additional perturbation can push boundary zeros slightly outside the closed disk; as written this point is implicit.","section":"Lemma 2.1"},{"comment":"The implicit-function-theorem argument for expressing partial Omega_{epsilon} as a normal graph is terse; for instance, the periodicity and C^k regularity of s(theta) and omega(theta) are asserted rather than shown. A few clarifying sentences would help.","section":"Corollary 2.2"},{"comment":"There are small presentation errors: Figure 2's caption contains the typo 'eignfunction', and Table 2's caption says 'Same as Figure (1)' where 'Table 1' is meant.","section":"Figures and tables"}],"recommendation":"major_revision","confidential_remarks":"The central idea and most of the proof are convincing, and the result is likely correct in the per-eigenfunction form suggested by the abstract. The main concerns are the quantifier in Theorem 1, the underspecified import from [GT19] in Lemma 4.1, and the apparent exponent typo in the proof of Lemma 4.1; all are fixable within the scope of the manuscript. The paper would be strengthened by stating the exact form of [GT19, Corollary 1.3] and by either proving or citing a published version with page/theorem number, especially because one of the authors is a co-author of [GT19]."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read it. The short version: this is a real result. It answers Girouard–Polterovich Open Problem 10(i) in the negative for a dense family of analytic domains: each has a fixed-size ball, depending only on the domain, where every Steklov eigenfunction is nonzero. That is new, and it breaks the heuristic that interior Steklov eigenfunctions behave like high-energy Laplace eigenfunctions.\n\nWhat I found strongest is the architecture. The tunneling condition is a fresh definition, and Theorem 3 proves it for BBLCN domains by a clean recurrence on Fourier coefficients; Lemma 3.1 is self-contained and transparent. Theorem 2 then turns exponential concentration of mass into a finite-mode approximation of the harmonic extension, which is exactly what is needed to force a zero-free ball. The paper is also honest about its limits: Remark 1.2 explicitly says the ellipse and kite numerics are not proven tunneling, and Conjecture 1.3 is clearly marked as a conjecture.\n\nThe one place I would push back on the stress-test note: the [GT19, Corollary 1.3] import does not look circular. It is a published, parameter-free estimate about Steklov eigenfunction Fourier coefficients, used as a lemma, not derived from the target theorem. The concern that it may be weaker than stated is fair as a verification burden, but my reading is that it is exactly a lower bound on low-frequency mass in a window around m. Still, the point is load-bearing: Lemma 4.1 bridges the tunneling definition to the exponential lower bound, and Theorem 2 collapses without it. A referee should verify the exact statement and the m=0 case.\n\nMinor soft spots: the m=0 use in Lemma 4.1 needs the [GT19] constant to be uniform in m, which the text asserts but does not show; and Corollary 2.2's implicit function theorem step is sketched rather than fully detailed. Both are repairable without touching the central idea.\n\nOverall this deserves a serious referee and likely acceptance after the external lemma is checked. The numerical section is supporting, not essential, and the analytic argument stands on its own.","headline":"A genuinely new negative answer to a named Steklov open problem, with a coherent proof built on one load-bearing imported estimate that deserves referee scrutiny.","tokens_in":18987,"tokens_out":1476,"would_cite":true,"duration_ms":15996,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35P15","35J05","35B05","30C35"],"pacs":[],"model":"deepseek-v4-flash","headline":"Steklov eigenfunctions on a dense family of analytic domains can remain nonzero on a fixed-radius ball, so their nodal sets are not dense at shrinking scales.","keywords":["Steklov problem","nodal sets","tunneling condition","BBLCN domains","conformal mapping","boundary Fourier coefficients","high-frequency asymptotics","spectral geometry"],"falsifier":"Compute boundary Fourier coefficients of high Steklov eigenfunctions on a non-BBLCN analytic domain, such as the ellipse with semiaxes 2 and 1; if for some sequence $\\sigma_j\\to\\infty$ the mass in the window $|k-m|\\le 2\\sigma_j$ falls below $(1-Ce^{-\\sigma_j/C})$ of the total mass, the exponential-concentration input to Lemma 4.1 fails and the proof chain collapses. A direct check of the paper's conclusion would be to search for a high eigenfunction whose zero set intersects every ball of radius $r_1$ inside the domain for every $r_1>0$.","tokens_in":17977,"feed_emoji":"📐","tokens_out":12208,"duration_ms":118658,"temperature":0.7,"pith_summary":"This paper answers an open question about Steklov eigenfunctions: whether their zero sets must become dense near the high-frequency limit. It shows the answer is no. For any bounded simply-connected planar domain with analytic boundary, one can perturb the boundary by an arbitrarily small analytic deformation so that, in the new domain, every Steklov eigenfunction is nonzero on some ball of a fixed radius, with the location of the ball allowed to depend on the eigenvalue. Therefore, for these domains, the nodal set is not dense at scale $\\sigma^{-1}$, and in a fixed interior region the eigenfunctions oscillate no faster than a bounded frequency. The proof works by showing that a dense class of domains, those with boundary-band-limited conformal maps, satisfy a tunneling condition forcing the interior eigenfunction to be dominated by finitely many Fourier modes.","feed_headline":"Steklov zeros can avoid a fixed-size ball","feed_subtitle":"Near any analytic domain, a tiny boundary perturbation makes every high eigenfunction nonzero on a ball of fixed radius.","key_machinery":"The central device is the tunneling condition, defined via boundary Fourier coefficients of the pulled-back eigenfunction $u=\\varphi\\circ f$ on the unit disk: for any $K$ there is $C_0$ such that $|\\hat u(k)|\\le C_0^{|k-m|}A_m$ for $|k|\\le K\\sigma$, where $A_m=(\\sum_{k=m-m_0}^{m+m_0}|\\hat u(k)|^2)^{1/2}$. Lemma 4.1 turns this into a lower bound $e^{-C\\sigma}\\|\\hat u\\|_{\\ell^2}\\le A_m$ on the low-frequency mass, so the interior harmonic extension cannot become exponentially negligible. Theorem 3 proves the tunneling condition for BBLCN domains, i.e. domains whose conformal mapping $f$ has $|\\partial_z f|$ boundary-band-limited and nonconstant on $\\partial\\mathbb{D}$, for example $f(z)=\\int p(w)^2\\,dw$ with polynomial $p$ having no roots in the disk. Corollary 2.2 approximates any analytic simply-connected domain arbitrarily closely in $C^k$ by such domains. Theorem 2 then uses the $r^{|k|}$ decay of Fourier harmonics to dominate high modes, leaving a low-frequency polynomial that cannot vanish throughout any fixed-radius ball.","core_discovery":"The paper's central discovery is that high-frequency Steklov eigenfunctions can be frozen inside a domain. Theorem 1 states that, starting from any bounded simply-connected analytic domain $\\Omega_0$, one can make an arbitrarily small analytic boundary perturbation (of size $\\varepsilon$ in $C^k$) to obtain a domain $\\Omega_1$ with a point $x_0$ and radii $0<r_1<r_0$ such that every Steklov eigenfunction of every eigenvalue $\\sigma$ is nonzero on some ball $B(x_\\sigma,r_1)$ inside $B(x_0,r_0)$. The point $x_\\sigma$ may depend on $\\sigma$, but $r_1$ does not. Since the expected oscillation scale is $\\sigma^{-1}$, this is incompatible with the nodal set being dense at that scale, giving a negative answer to Open Problem 10(i) of [GP17]. The refined Theorem 2 shows that on a small ball the eigenfunction is approximated in $C^N$ by finitely many Fourier modes, with relative error $C_N(\\delta m^{-N-m_0-1}+e^{-c\\sigma})$, and Theorem 3 identifies a dense class of domains, the BBLCN domains, for which the necessary tunneling estimate holds.","pith_inferences":["If the tunneling conjecture stated in the paper is true, the fixed-radius sign-ball behavior would extend to every analytic non-circular domain, making non-density generic; the ellipse and kite numerics already show the expected opening.","The proof reframes interior Steklov decay as an exponential tunneling effect, suggesting that interior Steklov eigenfunctions localize away from some regions in a way high-energy Laplace eigenfunctions do not; a comparison of nodal length asymptotics on the same domain could test this distinction.","A parameter-free extension would be to run the BBLCN construction with rational maps instead of polynomial squares, allowing the method to reach multiply connected or higher-genus geometries once the band-limited algebra is verified.","A quick numerical check of the mechanism is to perturb the unit disk by one low-order Fourier mode and track the largest sign-definite ball as $\\sigma$ grows; the theory predicts the radius stays bounded below, whereas for the exact disk it shrinks like $\\sigma^{-1}$."],"forward_implications":["Open Problem 10(i) of [GP17] has a negative answer: in the constructed domains, Steklov nodal sets are not dense at scale $\\sigma^{-1}$, and a fixed-radius ball of constant sign exists for every eigenfunction.","The class of simply-connected analytic domains exhibiting this behavior is dense in the $C^k$ boundary topology, so the phenomenon is not confined to special shapes.","On such domains, interior Steklov eigenfunctions are well approximated by finitely many Fourier modes: the $C^N$ error on a small ball is bounded by $C_N(\\delta m^{-N-m_0-1}+e^{-c\\sigma})$ relative to the $L^2$ norm.","The eigenfunctions have bounded frequency on a fixed interior neighborhood, so their oscillation rate does not grow with $\\sigma$ in that region.","Every analytic simply-connected domain can be approximated arbitrarily closely by a BBLCN domain, and every BBLCN domain is tunneling."],"supporting_citations":[{"why":"Raises Open Problem 10(i) on density of Steklov nodal sets at scale $\\sigma^{-1}$; Theorem 1 answers it negatively.","marker":"[GP17]"},{"why":"Supplies the exponential concentration of boundary Fourier coefficients in the window $|k-m|\\le 2\\sigma$, used in Lemma 4.1 to obtain the low-frequency mass lower bound.","marker":"[GT19]"},{"why":"Gives the smooth extension of conformal maps to the boundary, used to approximate $\\partial_z f$ by polynomials and construct BBLCN approximating domains.","marker":"[BK87]"},{"why":"Establishes the discrete Steklov spectrum that frames the eigenvalue problem and the sequence of eigenfunctions under study.","marker":"[HL01]"}],"fun_headline_variants":["Steklov modes each leave a ball zero-free","Every Steklov eigenfunction has a zero-free ball","Steklov nodal sets not dense: a ball avoids zeros","Fixed-radius ball with no Steklov zeros","Zeros dodge a ball in Steklov eigenfunctions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole chain depends on an imported exponential concentration estimate for the boundary Fourier coefficients of a Steklov eigenfunction; if that estimate fails for some high eigenvalue, the low-frequency mass bound and the resulting constant-sign ball are not established.","fun_headline_variants_meta":{"raw":{"variants":["Steklov modes each leave a ball zero-free","Every Steklov eigenfunction has a zero-free ball","Steklov nodal sets not dense: a ball avoids zeros","Fixed-radius ball with no Steklov zeros","Zeros dodge a ball in Steklov eigenfunctions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000543,"raw_usage":{"total_tokens":2663,"prompt_tokens":1072,"completion_tokens":1591,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":688,"completion_tokens_details":{"reasoning_tokens":1511}},"tokens_in":688,"tokens_out":1591,"duration_ms":16767,"temperature":1.0,"reasoning_tokens":1511,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:18:00.119919+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute boundary Fourier coefficients of high Steklov eigenfunctions on a non-BBLCN analytic domain, such as the ellipse with semiaxes 2 and 1; if for some sequence $\\sigma_j\\to\\infty$ the mass in the window $|k-m|\\le 2\\sigma_j$ falls below $(1-Ce^{-\\sigma_j/C})$ of the total mass, the exponential-concentration input to Lemma 4.1 fails and the proof chain collapses. A direct check of the paper's conclusion would be to search for a high eigenfunction whose zero set intersects every ball of radius $r_1$ inside the domain for every $r_1>0$.","supporting_citations":[],"review_version":1}