{"id":"d10ec9b5-bfe0-465e-a5e2-3c55f7e4ac36","arxiv_id":"1908.03256","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For bounded pseudoconvex domains in C^n, the variational eigenvalues of the ∂-Neumann Laplacian are upper and lower semicontinuous under Hausdorff perturbation, with quantitative Lipschitz-type rates on uniformly finite D'Angelo type domains.","lead":"This paper proves that the eigenvalues of a central operator in complex analysis, the ∂-Neumann Laplacian, change controllably when the underlying domain is slightly deformed. A generalist might read it to see how a non-coercive boundary value problem behaves under domain perturbations, with explicit rates for finite-type domains.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.3's two-sided linear rate relies on the first inequality in (5.22), whose proof is explicitly omitted ('left to the interested reader'); until that reverse direction is independently verified, the central claim is not fully established.","rationale":"The reader's strongest claim and weakest assumption correctly identify uniform finite D_q-type as the key input for the quantitative linear rate. My stress-test pass confirms that the omitted lower-bound proof is the most load-bearing unresolved point: Theorem 1.3 states a two-sided estimate, but the manuscript proves only one side in detail and leaves the reverse direction to the reader. This is a proof-completeness concern rather than a defect in the mathematical strategy; the construction is symmetric enough that the gap is likely fillable under the stated hypotheses. I therefore do not move the verdict away from CONDITIONAL, but I would ask the authors to supply the missing argument before final acceptance. The paper's other components—upper semicontinuity (Theorem 3.5), lower semicontinuity under property (P) (Theorem 4.4), resolvent convergence (Theorem 6.1)—are developed with substantial detail and no clear internal inconsistency. The broken sentence in Remark 3 and the omitted proof of (3.18) are minor and not load-bearing. My agreement with the reader is partial because the reader's weakest_assumption emphasizes uniform finite type as an input condition, whereas I would emphasize the omitted reverse direction in Theorem 5.6 as the concrete point that must be verified; both are legitimate, but the missing proof is the more actionable concern.","tokens_in":88,"tokens_out":8348,"duration_ms":383179,"concrete_test":"Independently derive the first inequality in (5.22) by reversing the roles of Ω and Ω_j in the proof of Theorem 5.6: construct a transition operator from Dom(Q_{Ω_j}) to Dom(Q_Ω) using the push-in/pull-back construction of (5.12)–(5.14) with Ω and Ω_j swapped, and check whether the analogues of (5.34)–(5.47) hold with the same powers of δ_j, using the uniform finite D_q-type constants for {Ω_j}. If the reverse argument requires an additional uniform bound (such as uniform C^2 bounds on the defining functions of Ω_j or a uniform constant for the extension operator E_j), then Theorem 1.3's hypothesis is insufficient as stated; if it goes through unchanged, the gap is expository and the theorem is sound.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central quantitative claim, Theorem 1.3, is exactly the two-sided Lipschitz bound |λ_k^q(Ω_j) - λ_k^q(Ω)| ≤ C_k δ_j. The proof of Theorem 5.6 constructs a transition operator T_j from Dom(Q_Ω) to Dom(Q_{Ω_j}) and verifies the norm and form estimates needed for the upper bound λ_k^q(Ω_j) - λ_k^q(Ω) ≤ C_k δ_j, culminating in (5.44) and (5.47). The reverse inequality, which is the first inequality in (5.22), is dispatched with: 'The first inequality of (5.22) is proved similarly and is left to the interested reader.' This is not a minor detail: the lower bound requires a transition operator in the opposite direction, and the estimates must be uniform with respect to j. The proof would need to bound eigenforms of Ω_j in C^1 and control their normal components using the uniform finite D_q-type constants for the family Ω_j. While this is plausible, it is not written, and the manuscript gives no indication of whether an additional uniform hypothesis (e.g., uniform C^2 bounds on the defining functions of Ω_j or a uniform extension operator constant) is needed. The same pattern appears in Theorem 5.4, where the second inequality is also left to the reader. Because the central theorem's headline is the symmetric linear rate, an unproved half of that rate is a load-bearing gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper develops spectral stability theory for the variational eigenvalues of the ∂-Neumann Laplacian when the underlying bounded pseudoconvex domain is perturbed in Hausdorff distance. The main results are: (1) upper semicontinuity of λ_k^q under arbitrary pseudoconvex perturbations of a domain with C^1 boundary (Theorem 1.1); (2) lower semicontinuity for targets satisfying Catlin's property (P_{q−1}) (Theorem 1.2); (3) quantitative two-sided estimates on finite-type domains, culminating in the linear Lipschitz bound |λ_k^q(Ω_j)−λ_k^q(Ω)| ≤ C_k δ_j under uniform finite D_q-type (Theorem 1.3, from (5.22)); and (4) strong resolvent convergence of the ∂-Neumann Laplacians (Theorem 6.1). The proofs compare quadratic forms through transition operators built from Straube-type regularization, normal shifts, extension operators, and apply the min-max perturbation lemma (Lemma 2.1).","tokens_in":27991,"tokens_out":45249,"duration_ms":476727,"significance":"The quantitative stability theorem is the paper's central contribution. A linear-in-δ bound for variational eigenvalues of a non-coercive boundary value problem under domain perturbation would be a substantial advance, and Remark 4 shows the linear rate is sharp for balls. The paper is largely self-contained in its use of established tools (Hörmander's L2 estimates, Catlin's subelliptic theory, Straube's regularization, Brezis–Marcus and Davies estimates) and contains no fitted parameters or circularity. The upper semicontinuity and resolvent convergence parts are proved in detail. However, the two-sided quantitative conclusions rest on reverse inequalities whose proofs are omitted; until those are supplied, the headline theorem is not fully established.","major_comments":[{"comment":"The proof of Theorem 5.6 establishes only the second inequality in (5.22), i.e. λ_k^q(Ω_j)−λ_k^q(Ω) ≤ C_kδ_j, and the first inequality is dismissed with 'proved similarly and is left to the interested reader.' This first inequality is exactly the lower half of the two-sided bound in Theorem 1.3, so the omission is load-bearing. The reverse direction requires a transition operator mapping Dom(Q_{Ω_j}) into Dom(Q_Ω), uniform boundary-layer estimates for eigenforms of Ω_j obtained from Lemma 5.2 and Lemma 5.5 under the uniform finite-type hypothesis, and a uniform extension operator for the family Ω_j. None of these estimates are written. In addition, the application of Lemma 2.1 requires estimates on k-dimensional orthonormal sets and their cross terms; the manuscript verifies estimates only for a single normalized eigenform of Ω. Please complete this direction with full uniform constants.","section":"Theorem 5.6, (5.22); Theorem 1.3"},{"comment":"The proof of Theorem 5.4 states 'The proof of the other inequality in Theorem 5.4 is similar and is left to the interested reader.' Since (5.11) is presented as a two-sided quantitative estimate, and the missing direction is an upper bound with rate δ^{α/(α+1)} that is not a formal consequence of the qualitative Theorem 3.5, this is not a cosmetic omission. The reverse transition from eigenforms of Ω to test forms on Ω_j must be supplied, including the boundary-layer estimates corresponding to (5.15)–(5.16) on the Ω_j side. As written, Theorem 5.4 is only half-proved.","section":"Theorem 5.4, (5.11)"},{"comment":"The hypothesis 'Suppose Ω_j and Ω are of uniform finite D_q-type' is ambiguous because uniformity is defined only for a family of domains with a common constant C in (5.2) and uniformly bounded defining functions. To make the linear rate in (5.22) meaningful, the statement must require that the enlarged family {Ω_j} ∪ {Ω} satisfies these uniform bounds. If Ω is not included in the uniform family, the constant C_k in (1.3) could a priori depend on j, and the uniform estimates from Lemma 5.5 used to improve δ^{1/2} to δ would fail. Please restate the theorem with the family explicitly containing Ω and with the uniform constants identified.","section":"Theorem 1.3 and the definition before Lemma 5.5"}],"minor_comments":[{"comment":"The sentence 'Since Ω does not have Stein neighborhood basis ([DF77b]), we have This follows from the fact that...' is grammatically incomplete and does not convey the intended argument. Please rewrite the remark.","section":"Remark 3, Section 4"},{"comment":"The function b is constructed only on a neighborhood U of ∂Ω, but Lemma 4.1 is applied on all of Ω_j. Please add a sentence or a citation explaining the standard extension of b to a bounded plurisubharmonic function on Ω_j with the same Hessian lower bound on the boundary collar.","section":"Lemma 4.2"},{"comment":"The proofs of these two lemmas are omitted. A one-sentence indication that they follow from Lemma 4.2 with δ^{2α} in place of ε^2 would help the reader verify the uniformity of the constants.","section":"Lemma 5.2 and Lemma 5.1"},{"comment":"The estimates in the proof are written for one normalized eigenform f of Ω. The step to Lemma 2.1 requires estimates for an orthonormal k-tuple and for the inner products ⟨T_j f_h, T_j f_l⟩−δ_{hl}; the polarization argument and the dependence of the constants on λ_k(Ω) should be stated explicitly.","section":"Theorem 5.6 proof"},{"comment":"The improvement from δ^{1/2} to δ for the cross term in the uniform finite-type case relies on the volume bound |A_{j,3δ_j}| = O(δ_j) and uniform C^0 bounds on f_{δ_j}−Ef, which follow from the uniform C^∞ bounds on the defining functions but are not stated. Please add a sentence.","section":"Proof of Theorem 5.6, (5.43)"},{"comment":"The phrase 'N_j does not converges to N in norm' should read 'N_j does not converge to N in norm'.","section":"Section 6, Remark 5(2)"},{"comment":"The phrase 'the C∞-norm of ρ_j is uniformly bounded' should be quantified as uniform bounds on all derivatives of ρ_j up to each order; the singular 'C∞-norm' is imprecise.","section":"Definition before Lemma 5.5"}],"recommendation":"major_revision","confidential_remarks":"The two-sided quantitative claims are the advertised headline of the paper. The omitted reverse inequalities in Theorem 5.4 and especially Theorem 5.6 are, in my view, fillable using the same machinery, but they are load-bearing and should be required rather than left to the reader's goodwill. I also recommend that the editors ask for a precise statement of the uniform finite-type hypothesis including Ω in the family."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is the first stability theory for variational eigenvalues of the ∂-Neumann Laplacian under Hausdorff domain perturbation, and the main results look right in outline. The paper proves upper semicontinuity on bounded pseudoconvex domains (C1 target), lower semicontinuity assuming property (P), and quantitative rates on finite type domains, including a linear rate under uniform finite D_q-type. That is a genuine step forward; the existing literature covers classical Laplacians and spectral characterizations of ∂-Neumann geometry, not domain perturbation of the eigenvalues themselves.\n\nThe proof strategy is sound: Straube's regularization for restriction, push-out/push-in of forms along normals for enlargement/shrinking, Catlin's barrier potentials for boundary estimates, and Brezis-Marcus Hardy plus Davies for the quantitative part. The basic comparison lemma is standard and the applications are mostly detailed. I found no circular reasoning or fitted parameters, and the authors are not overselling the novelty.\n\nNow the soft spots, in proportion. The central quantitative claim, Theorem 1.3, is the two-sided bound |λ_k^q(Ω_j) − λ_k^q(Ω)| ≤ C δ_j. The proof of Theorem 5.6 explicitly constructs the transition operator and proves one direction (the upper bound in (5.22)), then says the first inequality is proved similarly and left to the interested reader. That is not cosmetic: the reverse bound needs a transition in the opposite direction and uniform estimates under the same finite-type hypothesis. It is plausible, and maybe routine given the uniform assumption, but it is not written. The same pattern occurs in Theorem 5.4, where the second inequality is left to the reader. Until that reverse half is written, the headline linear rate is not fully established.\n\nSeveral auxiliary lemmas (3.4 second half, 4.2, 5.1, 5.2) are stated with proofs omitted or sketched; these look standard, so minor. Remark 3 on worm domains contains an incomplete sentence (\"Since Ω does not have Stein neighborhood basis, we have This follows from...\") — clearly a corrupted passage, needs fixing.\n\nWho this is for: SCV and spectral/domain-perturbation people. It deserves a serious referee. I would send it out, but the referee must be told to verify the omitted direction in 5.22 and the uniform constants. If that direction fails, the paper still has value but the headline changes.","headline":"First stability theory for ∂-Neumann variational eigenvalues under Hausdorff perturbation; sound and novel, but the proof of one half of the central linear-rate theorem is omitted, so the headline claim is not fully verified.","tokens_in":28525,"tokens_out":1814,"would_cite":true,"duration_ms":19368,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["32W05","32G05","35J25","35P15"],"pacs":[],"model":"deepseek-v4-flash","headline":"The variational eigenvalues of the ∂-Neumann Laplacian are Lipschitz stable under Hausdorff perturbation of smooth bounded pseudoconvex domains of uniform finite D'Angelo type.","keywords":["∂-Neumann Laplacian","spectral stability","variational eigenvalues","pseudoconvex domain","property (P)","finite type condition","Hausdorff distance","resolvent convergence"],"falsifier":"Compute, for a family of smooth bounded pseudoconvex domains of uniform finite $D_q$-type converging in Hausdorff distance to a limit domain $\\Omega$, the ratio $|\\lambda_1^q(\\Omega_j) - \\lambda_1^q(\\Omega)| / d_H(\\Omega_j, \\Omega)$. If this ratio is unbounded as $j \\to \\infty$, Theorem 1.3 is false. A concrete place to look is a family of smoothly flattened ellipsoids whose defining functions are renormalized to keep all $C^\\infty$ norms and subellipticity constants bounded; the ball computation of Remark 4 shows the calculation is feasible and the ratio is nonzero there.","tokens_in":27505,"feed_emoji":"📐","tokens_out":22070,"duration_ms":206638,"temperature":0.7,"pith_summary":"This paper asks what happens to the spectrum of the $\\bar\\partial$-Neumann Laplacian, the basic self-adjoint operator behind $L^2$ methods in several complex variables, when the underlying domain in $\\mathbb{C}^n$ is slightly deformed. It proves that for bounded pseudoconvex domains the variational eigenvalues are upper semi-continuous in the Hausdorff distance: a small perturbation cannot push any eigenvalue upward by more than an arbitrarily small amount. With the additional potential-theoretic assumption known as property (P), it proves the matching lower semi-continuity, so the eigenvalues actually converge under domain perturbation. The main quantitative result is that on smooth bounded pseudoconvex domains of uniform finite D'Angelo type, the $k$-th variational eigenvalue changes by at most a constant $C_k$ times the Hausdorff distance between the two domains—a linear, Lipschitz rate that the authors show is sharp. This matters because exact $\\bar\\partial$-Neumann spectra are almost never computable; the result guarantees that approximating a domain in Hausdorff distance gives a controlled approximation of its spectrum.","feed_headline":"Domain perturbations shift ∂-Neumann eigenvalues at most linearly","feed_subtitle":"For finite-type pseudoconvex domains, approximating the domain gives provably close spectra—the linear rate is sharp.","key_machinery":"The central object is the $k$-th variational eigenvalue $\\lambda_k^q(\\Omega)$ of the $\\bar\\partial$-Neumann Laplacian, defined by the min-max principle over $k$-dimensional subspaces of the quadratic-form domain $\\mathrm{Dom}(Q_q)$. The load-bearing mechanism is the decomposition of a $(0,q)$-form into tangential and normal components at the boundary: the tangential component is handled like a classical Neumann problem and the normal component like a Dirichlet problem, which overcomes the non-coercive nature of the $\\bar\\partial$-Neumann boundary conditions. This decomposition is combined with the plurisubharmonic barrier construction of finite-type theory (which yields subelliptic estimates and boundary decay for eigenforms) and with a sharp Hardy inequality and boundary estimates for elliptic operators to produce the quantitative eigenvalue comparison. A general spectral-stability lemma (Lemma 2.1) converts closeness of the transported quadratic forms into a bound on the variational eigenvalues.","core_discovery":"The central discovery is quantitative spectral stability in the finite-type regime. For smooth bounded pseudoconvex domains $\\Omega_j$ and $\\Omega$ in $\\mathbb{C}^n$ that are of uniform finite $D_q$-type, $1 \\le q \\le n-1$, the paper establishes that for every positive integer $k$ there exist constants $\\delta > 0$ and $C_k > 0$ such that $|\\lambda_k^q(\\Omega_j) - \\lambda_k^q(\\Omega)| \\le C_k \\, d_H(\\Omega, \\Omega_j)$ whenever the Hausdorff distance $d_H(\\Omega, \\Omega_j) < \\delta$. The linear dependence on $d_H$ is sharp: for the unit ball $B$, the scaling $\\lambda_k^q(rB) = r^{-2}\\lambda_k^q(B)$ gives a gap proportional to $d_H(rB, B)$, so no exponent better than $1$ is possible in general. The proof works by decomposing forms in the domain of the quadratic form into tangential and normal components—the tangential part behaves like a Neumann problem and the normal part like a Dirichlet problem—and then controlling both with the plurisubharmonic barrier construction of finite-type theory, a sharp Hardy inequality, and boundary estimates for elliptic operators.","pith_inferences":["The tangential/normal decomposition strategy is likely to transfer to other non-coercive boundary-value problems whose quadratic forms respect a similar splitting (for instance, the Kohn Laplacian on CR manifolds with boundary), giving analogous stability statements that the authors do not state.","The proof shows that the key input is not literally D'Angelo type but the uniform subellipticity estimate (5.2); any family of domains with uniformly controlled subellipticity constants and defining-function norms would satisfy the same Lipschitz conclusion, so the theorem can be read as a spectral-stability result for uniform property $(P^\\alpha_q)$.","A testable prediction from the quantitative method: if the subellipticity exponent $\\alpha$ in the family degenerates to $0$, the optimal rate should degrade from $\\delta$ to $\\delta^{\\alpha/(\\alpha+1)}$ (the general bound in Theorem 5.4), so the linear rate is intrinsically tied to a uniform positive lower bound on $\\alpha$.","The strong resolvent convergence suggests that numerical or approximate computations of $\\bar\\partial$-Neumann spectra—for example, on smoothly approximated domains—are mathematically justified for pseudoconvex domains with $C^1$ boundaries, which could be useful in computational complex analysis."],"forward_implications":["For any bounded pseudoconvex domain with $C^1$ boundary, the variational eigenvalues cannot jump upward under a Hausdorff-small pseudoconvex perturbation (Theorem 1.1), and they converge if the limit satisfies property (P) (Corollary 4.5).","When the limit domain is smooth, bounded, pseudoconvex, and of finite $D_q$-type, an upper bound of the form $\\lambda_k^q(\\Omega_j) - \\lambda_k^q(\\Omega) \\le C_k \\sqrt{\\delta_j}$ holds even without uniform type assumptions on the approximating domains (Theorem 5.6).","The $\\bar\\partial$-Neumann Laplacian converges in the strong resolvent sense for any sequence of bounded pseudoconvex domains with $C^1$ boundaries converging in Hausdorff distance; norm resolvent convergence can fail when the boundary of the limit contains complex varieties (Theorem 6.1 and Remark 5).","The linear rate in the finite-type theorem is optimal: the ball family shows that the eigenvalue gap can be exactly proportional to the Hausdorff distance (Remark 4).","Together, the upper and lower semi-continuity results give full continuity of the variational eigenvalues for pseudoconvex domains satisfying property (P) on the relevant form levels, extending classical domain-perturbation results for the Dirichlet and Neumann Laplacians to the non-coercive setting."],"supporting_citations":[{"why":"Supplies the subelliptic estimate and the plurisubharmonic barrier construction (5.2) that underlie the finite-type quantitative analysis.","marker":"[Ca87]"},{"why":"Introduces property (P) and the Hessian lower-bound lemma (Lemma 4.1) used to obtain boundary decay estimates for eigenforms.","marker":"[Ca84]"},{"why":"Provides the regularization procedure that transports forms between nested domains and extends property (P)/subellipticity implications to nonsmooth boundaries.","marker":"[S97]"},{"why":"Supplies the sharp Hardy inequality for the weight $d(z)^{-2}$ that controls the normal component of eigenforms in the boundary layer.","marker":"[BM97]"},{"why":"Supplies sharp boundary estimates for elliptic operators that yield the $\\delta^{3/2}$ decay of the normal component in Lemma 5.8.","marker":"[D00]"},{"why":"Provides the general spectral-stability lemma (compared with Lemma 2.1) that converts quadratic-form closeness on transported subspaces into variational-eigenvalue estimates.","marker":"[BL07]"}],"fun_headline_variants":["Eigenvalues of ∂-Neumann shift at most linearly under domain moves","Sharp linear bound for ∂-Neumann eigenvalues under domain perturbation","Domain perturbations: ∂-Neumann spectra vary at most linearly","Eigenvalue gaps shrink linearly as domains approach in Hausdorff metric","Finite-type ∂-Neumann eigenvalues are Lipschitz in domain"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The linear rate assumes the entire family of perturbed domains shares uniform finite $D_q$-type—the same subellipticity constants and uniformly bounded $C^\\infty$ norms of the defining functions—so that the barrier, Hardy, and boundary estimates hold with constants independent of the domain.","fun_headline_variants_meta":{"raw":{"variants":["Eigenvalues of ∂-Neumann shift at most linearly under domain moves","Sharp linear bound for ∂-Neumann eigenvalues under domain perturbation","Domain perturbations: ∂-Neumann spectra vary at most linearly","Eigenvalue gaps shrink linearly as domains approach in Hausdorff metric","Finite-type ∂-Neumann eigenvalues are Lipschitz in domain"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000635,"raw_usage":{"total_tokens":2909,"prompt_tokens":908,"completion_tokens":2001,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":524,"completion_tokens_details":{"reasoning_tokens":1908}},"tokens_in":524,"tokens_out":2001,"duration_ms":13974,"temperature":1.0,"reasoning_tokens":1908,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:20:15.550065+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, for a family of smooth bounded pseudoconvex domains of uniform finite $D_q$-type converging in Hausdorff distance to a limit domain $\\Omega$, the ratio $|\\lambda_1^q(\\Omega_j) - \\lambda_1^q(\\Omega)| / d_H(\\Omega_j, \\Omega)$. If this ratio is unbounded as $j \\to \\infty$, Theorem 1.3 is false. A concrete place to look is a family of smoothly flattened ellipsoids whose defining functions are renormalized to keep all $C^\\infty$ norms and subellipticity constants bounded; the ball computation of Remark 4 shows the calculation is feasible and the ratio is nonzero there.","supporting_citations":[],"review_version":1}