{"id":"a37b3613-8ec1-408c-a9dc-5d51acdd26cd","arxiv_id":"2505.03690","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 magnetic field vanishing to finite order, the ground state energies scale like β^(2/(κ*+2)) for Dirichlet/Neumann and β^(1/(κ0+2)) for Dirichlet-to-Neumann, with one-term asymptotics under non-degeneracy conditions.","lead":"This paper proves how the lowest energy levels of a magnetic Schrödinger operator in a bounded region grow as the magnetic field strength becomes infinite, in the case where the field can vanish to a higher order at some points. It covers Dirichlet, Neumann and Dirichlet-to-Neumann conditions, and gives the first exact correction terms under extra technical conditions.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"One-term expansions (Thms 8.6, 8.8) are not fully proven: the lower-bound halves of Thms 5.7-5.9 are omitted, yet they feed directly into the boundary local approximations (Thms 6.2-6.4).","rationale":"The leading-order theorems (Thms 1.1 and 1.3) are well supported: the m-function bounds in Section 3 and the estimates in Section 4 form a coherent, essentially complete proof, and I found no circularity or counterexample there. The conditional expansion part is less secure. First, it depends on a cluster of uniform non-degeneracy and geometric conditions (7.8)-(7.9), (8.12)-(8.13), (7.10) that are natural but are not verified in any genuinely new higher-order example. Second, and more concretely, three lower-bound proofs in Thms 5.7-5.9 are explicitly omitted, and those lower bounds are necessary inputs for Thms 6.2-6.4, which in turn underlie Thms 8.6 and 8.8. The omissions look fillable, and the leading-order results are unaffected, so a conditional acceptance is appropriate: accept once the missing lower-bound arguments are supplied. The reader's ACCEPT is close, but the manuscript as written does not yet contain a complete proof of the one-term expansion claims that are central to its abstract.","tokens_in":38361,"tokens_out":60258,"duration_ms":543808,"concrete_test":"Complete the omitted lower-bound half of Thm 5.7: prove that for the rectangle E(R)={|x'|<R, -R M_R < x_d < R} with M_R=max_{|x'|<R}|∇φ(x')|, one has λ_D(A,E(R)) ≥ λ_D(A,R^d_+) - C(R^{κ+1}M_R + R^{-2}), with C independent of R and φ. Then adapt the same argument to µ_N and µ_DN in Thms 5.8-5.9. If the inequality can be proved, the gap is fillable and the verdict can remain accept. If it fails for some C^{1,1} φ with M_R→0, then the error estimates in Thms 6.2-6.4 and hence the expansions in Thms 8.6 and 8.8 would need revision.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The load-bearing weakness is not in the leading-order theorems (Thms 1.1, 1.3), which are supported by a complete chain of m-function inequalities in Sections 3-4. It lies in the derivation of the one-term expansions. The boundary local approximation theorems Thms 6.2, 6.3 and 6.4 rely on Thms 5.7, 5.8 and 5.9 respectively. In each of Thms 5.7-5.9, the lower-bound inequality is explicitly omitted. For example, the proof of Thm 5.7 states: 'The lower bound for λ_D(A,Ω∩Q(0,R)) may be established in a similar manner, using E(R). We omit the details.' This lower bound is not the mirror image of the upper bound: the comparison set E(R)={|x'|<R, -R M_R < x_d < R} extends below the graph and is not contained in the half-space, so domain monotonicity does not directly give the needed control; one must show that the extra negative-x_d region, where the homogeneous field could create an additional well, does not lower the Dirichlet energy by more than O(R^{κ+1}M_R + R^{-2}). For the Neumann and DtN cases (Thms 5.8, 5.9), the flattening changes the boundary measure and the boundary condition, so the claimed 'similar perturbation argument' is a nontrivial step. Since Thms 8.6 and 8.8 invoke Thms 6.2-6.4 for the Γ_2 and Γ_0 contributions, the one-term asymptotic expansions are not fully established as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the ground state energies of the magnetic Laplacian with Dirichlet, Neumann, and Dirichlet-to-Neumann boundary conditions in a bounded Lipschitz domain, in the strong-field limit \\beta\\to\\infty. It assumes that the magnetic field may vanish to finite, possibly high, order. The main leading-order results (Theorems 1.1 and 1.3) state that \\lambda_D,\\lambda_N \\asymp \\beta^{2/(\\kappa_*+2)} and \\lambda_{DN} \\asymp \\beta^{1/(\\kappa_0+2)} under minimal non-vanishing-to-infinite-order assumptions. Under additional uniform non-degeneracy and distance assumptions, Theorems 8.6 and 8.8 give one-term asymptotic expansions with remainder estimates, with coefficients expressed as ground state energies of model operators associated with Taylor polynomials of the field. The proofs combine quasimode upper bounds, operator lower bounds based on an m-function and uncertainty principle, localization arguments, and comparison with polynomial model problems in half-spaces.","tokens_in":38753,"tokens_out":7725,"duration_ms":69387,"significance":"If fully justified, the paper would provide a unified treatment of Dirichlet, Neumann, and Dirichlet-to-Neumann ground state energies for magnetic fields with arbitrary finite-order vanishing, recovering and extending several known results in the non-vanishing, discrete-well, and first-order vanishing cases. The leading-order theorems are proved with explicit constants that depend only on the dimension, the domain, and the vanishing order, and the model-operator coefficients are parameter-free. The m-function technique avoids localization error terms and works uniformly for all three boundary conditions. The paper also makes concrete new contributions for \\kappa_*\\ge 2 and for the DtN operator. However, the one-term expansion results rely on omitted lower-bound estimates in Section 5, so the paper is not yet complete as written.","major_comments":[{"comment":"The lower-bound halves of Theorems 5.7, 5.8, and 5.9 are omitted. For example, the proof of Theorem 5.7 states that the lower bound for \\lambda_D(A,\\Omega\\cap Q(0,R)) \"may be established in a similar manner, using E(R)\", and the proofs of Theorems 5.8 and 5.9 refer to \"a similar perturbation argument\" without details. These lower bounds are not mirror images of the upper bounds: the comparison set E(R) extends below the graph and is not contained in the half-space, so domain monotonicity does not directly give the required control, and for the Neumann and DtN cases the flattening map changes both the boundary measure and the boundary condition. Since Theorems 6.2–6.4 invoke Theorems 5.7–5.9, and Theorems 8.5, 8.6, and 8.8 in turn invoke Theorems 6.2–6.4, the one-term asymptotic expansions are not fully established as written. This gap does not affect Theorems 1.1 and 1.3, whose proofs are complete.","section":"§5, Theorems 5.7–5.9"},{"comment":"The lower bound for \\lambda_N in Theorem 8.5 is justified by \"a similar argument, using Remark 8.4 and Theorem 6.3\". Since Theorem 6.3 inherits the missing lower-bound estimate from Theorem 5.8, and Remark 8.4 relies only on monotonicity of \\mu_N, the Neumann half of the one-term expansion has the same gap as the Dirichlet case. A complete proof would need to supply the omitted lower-bound arguments in Section 5 before the expansion theorems can be considered fully proven.","section":"§8, Theorem 8.5"}],"minor_comments":[{"comment":"In the remainder exponent, the symbol \"k∗+4\" should be \"\\kappa_*+4\".","section":"§1, Eq. (1.15)"},{"comment":"The trace inequality line contains the typo \"Bx0,2r)\", which should read \"B(x_0,2r)\".","section":"§3, proof of Theorem 3.10"},{"comment":"The phrase \"and and\" should be reduced to \"and\".","section":"§5, proof of Proposition 5.1"},{"comment":"The claim that the estimates in (6.11) and (6.17) hold uniformly for y\\in\\partial\\Omega in the case \\Gamma_*=\\partial\\Omega needs a separate justification: if B_{12} vanishes identically on an open boundary arc, then \\nabla B_{12} is normal to \\partial\\Omega, so the invariant subspace V is tangent to the boundary and the quantity \\tau(y) in (6.10) is zero. Since the proof of Theorem 6.2 uses \\tau(y)>0, this case is not covered by the stated hypotheses.","section":"§9, Remark 9.11"}],"recommendation":"major_revision","confidential_remarks":"The leading-order theorems are complete and valuable, and the expansion results are plausible. The main issue is that the omitted lower-bound estimates in Section 5 are load-bearing for Theorems 8.6 and 8.8. The gaps appear repairable, so I recommend inviting a revision rather than rejecting."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The leading-order part of this paper is the real thing. Theorem 1.1 and Theorem 1.3 prove the two-sided bounds c β^{2/(κ*+2)} ≤ λ_N ≤ λ_D ≤ C β^{2/(κ*+2)} and c β^{1/(κ0+2)} ≤ λ_DN ≤ C β^{1/(κ0+2)} under only the \"no infinite-order vanishing\" hypothesis, for bounded Lipschitz domains. This genuinely removes the submanifold/discrete-zero-set conditions from earlier work, and the m-function lower bounds in Sections 3–4 are a clean, unified tool that handles Dirichlet, Neumann, and DtN in one sweep. The recovery of known cases in Section 9 is careful. The self-citations [30,31] are not a problem: the key inequality is reproved as Theorem 3.8.\n\nThe soft spot is the one-term expansions. Theorems 8.6 and 8.8 depend on the boundary local approximation theorems 6.2–6.4, which in turn depend on Theorems 5.7–5.9. In each of those three proofs the lower-bound half is explicitly omitted: \"may be established in a similar manner\" or \"by a similar perturbation argument.\" The stress-test note is right that this is not a routine mirror argument. For Theorem 5.7 the lower bound must handle the set E(R) = {|x'|<R, -R M_R < x_d < R}, which sticks out below the half-space. The homogeneous field could in principle create an extra well in that strip, so monotonicity does not give the bound; you need to prove the strip lowers λ_D by at most O(R^{κ+1}M_R + R^{-2}). That is a real estimate, not a cosmetic detail. For Theorems 5.8 and 5.9 the flattening changes the boundary measure and the boundary condition, so \"similar\" is doing even more work. Since Theorems 6.2–6.4 invoke these lower bounds, the expansions (1.15)–(1.16) are not fully established as written.\n\nI want to be clear about proportion: this is not a flaw in the leading-order results. Theorems 1.1 and 1.3 stand on their own, and the gap is an omission, not a contradiction in the argument. The expansions are probably true and probably fixable. But a referee should ask for the missing lower bounds to be written out before relying on them.\n\nRecommendation: send to peer review with a request that the author complete the proofs of Theorems 5.7–5.9 (or clearly state what additional assumptions make the lower bounds routine). If that is fixed, this becomes a strong paper for the spectral theory audience.","headline":"Strong leading-order theorems with a genuine proof gap in the one-term expansions; a referee should ask for the omitted lower bounds.","tokens_in":39275,"tokens_out":6675,"would_cite":true,"duration_ms":58919,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35P25"],"pacs":[],"model":"deepseek-v4-flash","headline":"Magnetic ground-state energies are fixed by the vanishing order of the field.","keywords":["magnetic Laplacian","vanishing magnetic field","ground state energy","Dirichlet-to-Neumann operator","asymptotic expansion","semiclassical analysis","Schrödinger operator","bounded domain"],"falsifier":"Compute the lowest Dirichlet eigenvalue of (D+βA)^2 on a bounded $C^{{1,1}}$ domain in $R^{2}$ with B_{12}(x,y)=$x^{2}$, for example using the gauge A(x,y)=(0, $x^{3}$/3) on the unit square. The theorem predicts λ_D(βA,Ω) $β^{{-1/2}}$ converges to a positive constant as β→∞. If instead λ_D $β^{{-1/2}}$ drifts to 0 or ∞, or if for a field with κ*=3 the exponent differs from 2/5, the central claim fails.","tokens_in":38133,"feed_emoji":"🧲","tokens_out":5982,"duration_ms":55940,"temperature":0.7,"pith_summary":"This paper proves that for a bounded Lipschitz domain, the ground state energies of the magnetic Laplacian with Dirichlet, Neumann, and Dirichlet-to-Neumann conditions grow at explicit powers of the field strength, determined solely by the maximal order at which the magnetic field vanishes. If the field does not vanish to infinite order anywhere in the domain, the bulk operators behave like $β^{{2/(κ*+2)}}$, where κ* is the maximal vanishing order; on the boundary, the Dirichlet-to-Neumann operator behaves like $β^{{1/(κ0+2)}}$. Under additional uniformity conditions on the leading Taylor coefficients at maximal vanishing points, the paper identifies the first nonzero constant in each expansion as a minimum of model eigenvalues on the whole space and on half-spaces. A unified set of operator inequalities replaces the usual localization and commutator arguments, so all three boundary value problems are treated by one mechanism.","feed_headline":"Magnetic energy grows with the field's vanishing order","feed_subtitle":"Dirichlet, Neumann, and Dirichlet-to-Neumann cases all obey explicit powers of β, with leading constants.","key_machinery":"The workhorse is the operator lower bound built from the function m(x,B), defined by 1/m(x,B) = sup{r>0 : max_Q(x,r)|B| ≤ 1/$r^{2}$}. The paper proves c∫_Ω m(x,B)^2|ψ|^2 ≤ ∫_Ω |(D+A)ψ|^2 and a boundary variant c∫_{∂Ω} m(x,B)|ψ|^2 ≤ ∫_{Ω_b} |(D+A)ψ|^2, replacing the usual localization and commutator devices and working uniformly for all three boundary conditions. For the asymptotic expansions, the argument fixes a point y of maximal vanishing order, takes the κ(y)-th Taylor polynomial of B, builds from it a homogeneous polynomial potential A_y, and compares the genuine eigenvalues with the model eigenvalues λ(A_y,R^d), λ_D(A_y,H_{n(y)}), λ_N(A_y,H_{n(y)}), and λ_DN(A_y,H_{n(y)}). Uniformity of these comparisons is controlled by the invariant subspace V_y of the Taylor polynomial, namely the largest subspace on which the top-order polynomial is translation-invariant, together with the non-degeneracy conditions (1.28)-(1.29).","core_discovery":"The central claim is that arbitrary finite-order vanishing magnetic fields give universal scaling laws: c $β^{{2/(κ*+2)}}$ ≤ λ_N(βA,Ω) ≤ λ_D(βA,Ω) ≤ C $β^{{2/(κ*+2)}}$ and c $β^{{1/(κ0+2)}}$ ≤ λ_DN(βA,Ω) ≤ C $β^{{1/(κ0+2)}}$ for large β. The sharper Theorems 8.6 and 8.8 add the first term: λ_D(βA,Ω) = Θ_D $β^{{2/(κ*+2)}}$ + O($β^{{1/(κ*+2)+1/(κ*+4)}}$), the analogous expansion for λ_N, and λ_DN(βA,Ω) = Θ_DN $β^{{1/(κ0+2)}}$ + O($β^{{1/(κ0+4)}}$), under the non-degeneracy conditions (1.28)-(1.29) and the distance controls (8.12)-(8.13). The constants Θ_D, Θ_N, Θ_DN are minima over the points of maximal vanishing order of the ground state energies of the homogeneous polynomial model potentials A_y on R^d or on the half-space with inward normal n(y). This unifies previously separate results for non-vanishing fields, discrete wells, and first-order vanishing in two dimensions.","pith_inferences":[],"forward_implications":["For any magnetic field that vanishes only to finite order, the high-field scaling is now pinned: the exponent depends only on the maximal vanishing order, not on the shape of the zero set.","The Dirichlet-to-Neumann operator, previously studied mainly for non-vanishing or constant fields, obeys the boundary analogue λ_DN(βA,Ω) ≈ β^{1/(κ0+2)}, with a one-term expansion under the same uniformity conditions.","Known results for non-vanishing fields, discrete wells, and first-order vanishing in two dimensions are recovered as special cases of one framework.","The leading constants in the expansions are explicit minima over model problems, so computing the prefactor reduces to solving homogeneous polynomial model operators on R^d and on half-spaces.","The same lower-bound inequalities are strong enough to support localization estimates for eigenfunctions, which the paper notes as a natural follow-up.","If the sharper expansions hold, then measuring the growth rate of λ_D, λ_N, or λ_DN in β gives a direct spectral way to read off the vanishing order of the magnetic field.","A testable numerical extension is to take d=2 with B_{12}(x,y)=x^2 (so κ*=2) on the unit square and finite-element compute the lowest Dirichlet eigenvalue for large β; the theorem predicts λ_D(βA,Ω) β^{-1/2} tends to a positive constant.","The uniform non-degeneracy conditions (1.28)-(1.29) are likely stronger than necessary; a plausible conjecture is that the one-term expansions persist under a weaker averaged version of the same conditions with the same constants Θ_D, Θ_N, Θ_DN."],"supporting_citations":[{"why":"Introduced the Dirichlet vanishing-field problem and proved the first-order case λ_D ≈ β^{2/3}, the starting point generalized here.","marker":"[25]"},{"why":"Proved the leading order λ_D(βA,Ω) ≈ β^{2/(κ*+2)} under submanifold or discrete zero-set assumptions, the result this paper extends to general bounded Lipschitz domains.","marker":"[18]"},{"why":"Identified the leading behavior of the magnetic Dirichlet-to-Neumann eigenvalues in the non-vanishing case, the boundary analogue this paper treats for finite-order vanishing.","marker":"[17]"},{"why":"Supplies the weight function m(x,B) that is the basis for the two operator lower bounds (1.13)-(1.14).","marker":"[30]"},{"why":"Provides the approach connecting the magnetic field to the magnetic potential that the lower-bound proof relies on.","marker":"[31]"},{"why":"Supplies the uncertainty-principle ingredient used inside the proof of the lower bounds.","marker":"[12]"},{"why":"Previous Neumann vanishing-case result for d=2 and κ*=1 with a one-term expansion, recovered by the unified theorems.","marker":"[24]"}],"fun_headline_variants":["Higher-order vanishing fields set sharp β exponents for magnetic energies","Magnetic Laplacian: leading energy terms depend on vanishing order","Unified asymptotics for Dirichlet, Neumann, and DN operators","Explicit powers of β for magnetic ground states in bounded domains","Vanishing order of field dictates magnetic energy growth"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"For the sharper one-term expansions, the load-bearing premise is that at every point where the field vanishes to maximal order, the top-order Taylor coefficients stay uniformly non-degenerate in all directions not in the field's invariant subspace, and that maximal points on the boundary can be reached by interior maximal points; the leading-order β-power results do not need these conditions.","fun_headline_variants_meta":{"raw":{"variants":["Higher-order vanishing fields set sharp β exponents for magnetic energies","Magnetic Laplacian: leading energy terms depend on vanishing order","Unified asymptotics for Dirichlet, Neumann, and DN operators","Explicit powers of β for magnetic ground states in bounded domains","Vanishing order of field dictates magnetic energy growth"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000418,"raw_usage":{"total_tokens":2152,"prompt_tokens":943,"completion_tokens":1209,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":559,"completion_tokens_details":{"reasoning_tokens":1125}},"tokens_in":559,"tokens_out":1209,"duration_ms":10971,"temperature":1.0,"reasoning_tokens":1125,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T23:45:14.286355+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the lowest Dirichlet eigenvalue of (D+βA)^2 on a bounded $C^{{1,1}}$ domain in $R^{2}$ with B_{12}(x,y)=$x^{2}$, for example using the gauge A(x,y)=(0, $x^{3}$/3) on the unit square. The theorem predicts λ_D(βA,Ω) $β^{{-1/2}}$ converges to a positive constant as β→∞. If instead λ_D $β^{{-1/2}}$ drifts to 0 or ∞, or if for a field with κ*=3 the exponent differs from 2/5, the central claim fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduced the Dirichlet vanishing-field problem and proved the first-order case λ_D ≈ β^{2/3}, the starting point generalized here."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proved the leading order λ_D(βA,Ω) ≈ β^{2/(κ*+2)} under submanifold or discrete zero-set assumptions, the result this paper extends to general bounded Lipschitz domains."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the weight function m(x,B) that is the basis for the two operator lower bounds (1.13)-(1.14)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the approach connecting the magnetic field to the magnetic potential that the lower-bound proof relies on."},{"cited_title":"Fefferman, The uncertainty principle , Bull","cited_arxiv_id":null,"evidence_quote":"Supplies the uncertainty-principle ingredient used inside the proof of the lower bounds."},{"cited_title":"Henri Poincar´ e19 (2018), no","cited_arxiv_id":null,"evidence_quote":"Previous Neumann vanishing-case result for d=2 and κ*=1 with a one-term expansion, recovered by the unified theorems."}],"review_version":1}