{"id":"338a22c7-a113-44ef-8505-b38f5e99678e","arxiv_id":"1908.03255","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The variational eigenvalues of the Kohn-Nirenberg regularized ∂-Neumann Laplacian converge to those of the ∂-Neumann Laplacian as the regularization tends to zero, with quantitative rates on finite-type pseudoconvex domains, and domain perturbations are controlled by C^2 closeness.","lead":"This paper proves quantitative spectral stability estimates for the Kohn-Nirenberg elliptic regularization of the ∂-Neumann Laplacian, both as the regularization parameter goes to zero and as the underlying domain is perturbed. The results give explicit error bounds for variational eigenvalues in terms of the regularization parameter and the C^2 distance between domains.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.2 is not proven as stated: the two-sided domain-perturbation bound depends on the omitted reverse inequality (4.23), whose proof is explicitly left to the reader; without it only a one-sided estimate follows.","rationale":"The reader's weakest_assumption identifies the same missing reverse inequality, so I agree. The qualitative convergence results (Theorem 1.1 first parts, Theorem 3.3) appear supported by standard arguments, and the one-sided construction in Theorem 4.3 is plausible. But the central novel quantitative domain-perturbation claim rests on an unproved reverse estimate and on potentially miscomputed t-exponents. This is a proof gap, not a refutation; it should be fixable, so the appropriate disposition remains conditional rather than reject or accept. I recommend no change to the reader's conditional verdict.","tokens_in":15582,"tokens_out":19816,"duration_ms":178972,"concrete_test":"Supply the complete proof of (4.23) by reversing Ω and Ω_j in §4, tracking every constant through Lemma 4.1, Lemma 4.2, and the transition-operator estimates (4.17)–(4.22); verify the constants remain independent of j under the uniform C^∞ bound on r_j and that the final t-exponent equals t^{2^{n+3}-1}. If the reverse proof cannot be completed with the stated exponent, Theorem 1.2 must be revised.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 4 proves only the forward direction of the domain-perturbation bound: Theorem 4.3 and (4.16) give λ^{t,q}_k(Ω_j) ≤ λ^{t,q}_k(Ω) + C_k δ_j / t^{2^{n+3}-1}. The reverse inequality (4.23), λ^{t,q}_k(Ω) ≤ λ^{t,q}_k(Ω_j)+..., is never established; the text immediately after Theorem 4.3 says 'The proof of (4.23) is similar... We leave the details to the interested reader.' This is load-bearing because Theorem 1.2 is the absolute-value estimate |λ^{t,q}_k(Ω_j)-λ^{t,q}_k(Ω)| ≤ ..., and without (4.23) the two-sided claim does not follow. The swap also requires uniform j-independence of all constants in Lemmas 4.1, 4.2, and in the transition-operator estimates when Ω and Ω_j exchange roles; this is asserted via the uniform C^∞ bound on r_j but not shown. In addition, the exponent bookkeeping in the auxiliary estimates is not internally consistent in the text as written (e.g., the proof of Lemma 4.2 invokes (4.1) with s=0 but obtains a different t-power than either displayed inequality), so the exact t-rate in (1.3) is not reliably supported.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies spectral stability of the Kohn-Nirenberg elliptic regularization □^t_q of the ∂-Neumann Laplacian on bounded domains in Cn. Theorem 1.1 claims strong resolvent convergence of □^t_q to □_q as t→0+ for bounded C^2 domains, norm resolvent convergence on smooth strongly pseudoconvex domains, and a quantitative eigenvalue rate O(t k (λ^q_k)^{2([1/(2α)]+1)}) on smooth bounded pseudoconvex domains of finite D'Angelo type. Theorem 1.2 claims a quantitative two-sided domain-perturbation bound |λ^{t,q}_k(Ω_j)-λ^{t,q}_k(Ω)| ≤ C_k δ_j / t^{2^{n+3}-1} for close smooth bounded pseudoconvex domains. The proofs use min-max characterizations, a transition operator between domains, elliptic regularity estimates, and resolvent arguments based on standard theorems of Hörmander, Catlin, and Kohn.","tokens_in":15845,"tokens_out":14279,"duration_ms":139347,"significance":"If the results hold, they provide explicit quantitative rates for spectral stability of the ∂-Neumann Laplacian under the canonical elliptic regularization and under C^2 domain perturbation, complementing the authors' earlier domain-perturbation work. The paper is free of fitted parameters and derives the estimates from standard external theorems and min-max arguments; the proof of Theorem 1.1 is largely self-contained and the resolvent-convergence arguments are natural. The significant caveat is that the quantitative domain-perturbation theorem rests on an omitted reverse inequality and on exponent bookkeeping that appears inconsistent as printed, so the central rate in Theorem 1.2 is conditional on supplying and correcting those details.","major_comments":[{"comment":"The reverse inequality (4.23) is load-bearing for Theorem 1.2, but it is not proved. The text states 'The proof of (4.23) is similar... We leave the details to the interested reader.' As written, only the forward estimate (4.16) is established in Section 4, so the two-sided bound in Theorem 1.2 does not follow. In addition, the assertion that all constants in Lemmas 4.1, 4.2, and Theorem 4.3 remain independent of j when Ω and Ω_j are swapped is not demonstrated; the uniform C∞ bound on r_j is mentioned, but the proof does not track the dependence of the extension operator, the boundary-chart constants, or the constants in Lemma 4.2. These details must be supplied before Theorem 1.2 is proved.","section":"§4, after Theorem 4.3 (Eq. (4.23))"},{"comment":"The exponent bookkeeping in the proof of Lemma 4.2 is internally inconsistent. The proof says that 'from (4.1) with s=0' one obtains ||u||_{W^2} ≤ C/t^{3/2}||□^t u||, whereas (4.1) with s=0 gives C/t; and it says that 'from (4.1) with s=2' one obtains C/t^6, whereas (4.1) with s=2 gives C/t^5. Moreover, the displayed pseudoconvex estimate (4.2) has exponent 3·2^s−1, which for s=0 and s=2 gives t^2 and t^11, again not matching the powers t^{3/2} and t^{15/2} used in the proof. Since Lemma 4.2 feeds directly into (4.16) and (1.3), the exact t-power in Theorem 1.2 is not reliably established as written.","section":"§4, proof of Lemma 4.2"},{"comment":"The estimate of the derivative difference ∂(T_j u^t_h−u^t_h) is not justified in the text. For the normal-component pieces, the difference contains u^t_{h,J}(z+χ_j(d(z))n_l)−u^t_{h,J}(z), whose pointwise bound requires the C^1 estimate from Lemma 4.2 together with control of χ'_j and the size of the shift; none of these steps is shown. The stated denominator t^{(2^{n+3}−1)/2} depends on exactly such a bound, and this estimate is needed for the energy comparison (4.22) and hence for the forward direction (4.16).","section":"§4, Eq. (4.21)"}],"minor_comments":[{"comment":"The notation t^{2n+3−1} (and its variants in (4.10), (4.11), (4.16), (4.21), and (4.22)) is ambiguous; from the proofs it appears that t^{2^{n+3}−1} is intended. Please clarify the notation consistently.","section":"Eq. (1.3) and Section 4"},{"comment":"There is a duplicated word: 'there there exists v' in the first paragraph of the proof. This and the typo '∂-Neumannn' in the proof of Proposition 2.1 should be corrected.","section":"§3, proof of Theorem 3.3"},{"comment":"The reference [D00] is missing page/article details, and the title of [S10] contains the typo 'Partical Differential Equations' instead of 'Partial Differential Equations'.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The main issue is not novelty or mathematical sensibility but completeness of proof. The omitted reverse inequality is explicitly acknowledged as left to the reader, so the editor should require the authors to supply it in full and to correct the exponent tracking before acceptance."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Punchline: the qualitative spectral stability under t→0 is solid, and the quantitative rate (1.2) is a genuine new result; the two-sided domain-perturbation theorem, however, is not proved as stated. The reverse inequality is explicitly left to the reader, and the t-exponent bookkeeping in Lemma 4.2 is inconsistent as written.\n\nWhat is new: Theorem 1.1 gives strong resolvent convergence of the Kohn–Nirenberg Laplacian to the ∂-Neumann Laplacian on bounded C^2 domains, norm resolvent convergence in the strongly pseudoconvex case, and the quantitative eigenvalue bound (1.2) with explicit dependence on k and the eigenvalue on finite-type pseudoconvex domains. The forward one-sided domain-perturbation estimate (4.16) is also new. The proof of Theorem 3.2 is clean: min-max plus Catlin's subelliptic estimates and the density of W^1∩Dom(∂*) in Dom(Q) is enough. No fitted parameters, no invented entities; self-citations to [FZ19] are to genuinely prior work and are not circular.\n\nSoft spots: Theorem 1.2 requires the reverse inequality (4.23). The text after Theorem 4.3 says the proof is similar and leaves details to the reader. That is load-bearing because (1.3) is an absolute value; without (4.23), only the one-sided bound (4.16) follows. The assertion that constants remain uniform in j when Ω and Ω_j are swapped is plausible given the uniform C^∞ bound on r_j, but it is asserted, not demonstrated. Also, the exponents in Lemma 4.2 do not line up: the proof invokes (4.1) with s=0 and gets t^{-3/2}; combining s=0 and s=2 with the displayed (4.1) and (4.2) gives different powers than the printed ones. The exact t-power in (1.3) is therefore not reliably established. These are fixable, but they are central to the paper's advertised quantitative claim.\n\nBottom line: this deserves a serious referee, not a desk reject. I would send it out with instructions that the reverse estimate and the exponent arithmetic must be repaired, or the theorem restated to match what is proved. The qualitative results and the forward direction are good enough that the core ideas are likely sound. I would not cite (1.3) as it stands, but I might cite Theorem 3.2 and the resolvent convergence after the revisions.","headline":"Qualitative stability under t→0 is solid and the quantitative rate is new, but the two-sided domain-perturbation theorem is not proved as stated because the reverse inequality is omitted and the t-exponent bookkeeping in Lemma 4.2 is inconsistent.","tokens_in":16431,"tokens_out":5248,"would_cite":true,"duration_ms":49216,"reading_group":"maybe","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":"This paper proves that the Kohn-Nirenberg elliptic regularization of the ∂-Neumann Laplacian converges to the original operator in strong resolvent sense as t→0+, and that on finite-type pseudoconvex domains the eigenvalue error obeys an…","keywords":["∂-Neumann Laplacian","Kohn-Nirenberg elliptic regularization","variational eigenvalue","pseudoconvex domain","finite type condition","spectral stability","resolvent convergence","subelliptic estimate"],"falsifier":"Compute the low eigenvalues of $\\square^t_q$ on a domain with explicitly known $\\bar\\partial$-Neumann spectrum, such as the unit ball in $\\mathbb{C}^n$, for a decreasing sequence of $t$; the quantitative part of Theorem 1.1 requires $|\\lambda^{t,q}_1-\\lambda^q_1|=O(t)$, so observing decay slower than a constant times $t$ would falsify it. For Theorem 1.2, build a family $\\Omega_j$ converging to $\\Omega$ in $C^2$ and check whether $|\\lambda^{t,q}_1(\\Omega_j)-\\lambda^{t,q}_1(\\Omega)|/\\delta_j$ stays bounded by $C/t^{2n+3-1}$ as $\\delta_j\\to 0$; unboundedness would disprove the bound.","tokens_in":15332,"feed_emoji":"📐","tokens_out":17779,"duration_ms":153333,"temperature":0.7,"pith_summary":"The $\\bar\\partial$-Neumann Laplacian is a basic boundary-value problem of several complex variables, but its boundary condition is not coercive, so standard elliptic tools do not apply directly. This paper studies the Kohn-Nirenberg elliptic regularization, which adds a small multiple $t$ of the full gradient to the quadratic form; the regularized operator is coercive and has purely discrete spectrum. The paper proves that this regularized operator is spectrally close to the $\\bar\\partial$-Neumann Laplacian: as $t\\to 0^+$, the operators converge in strong resolvent sense and each variational eigenvalue converges. Under the finite-type condition, the convergence is quantitative, with an explicit power of $t$ and of the eigenvalue in the error bound. The paper also proves a quantitative stability estimate for the regularized eigenvalues when the domain itself is perturbed in the $C^2$ norm.","feed_headline":"Elliptic regularization converges to the ∂-Neumann spectrum","feed_subtitle":"Adding a tiny gradient term keeps the spectrum provably close as t→0, with explicit rates.","key_machinery":"The central object is the Kohn-Nirenberg Laplacian $\\square^t_q$, the self-adjoint operator associated with the quadratic form $Q^t_q(u,v)=Q_q(u,v)+t\\langle \\nabla u,\\nabla v\\rangle$ on $W^1_{(0,q)}(\\Omega)\\cap\\operatorname{Dom}(\\bar\\partial^*_{q-1})$; adding the $t$-gradient term makes the boundary value problem coercive and elliptic. The argument compares variational eigenvalues through the min-max principle, using a transition operator $T$ that maps forms on one domain to forms on a nearby domain by decomposing into tangential and normal parts. The quantitative parts rest on subelliptic estimates, which give Sobolev bounds on eigenforms of the form $\\|u\\|_{W^{2m\\alpha}}\\le B_m(\\lambda(\\Omega))^m\\|u\\|$, and on elliptic regularity estimates for $\\square^t_q$ whose constants the paper tracks as explicit powers of $t$.","core_discovery":"For a bounded $C^2$ domain $\\Omega\\subset\\mathbb{C}^n$, the paper establishes that $\\square^t_q\\to\\square_q$ in strong resolvent sense as $t\\to 0^+$ and $\\lim_{t\\to 0^+}\\lambda^{t,q}_k(\\Omega)=\\lambda^q_k(\\Omega)$ for every $k\\in\\mathbb{N}$ and $1\\le q\\le n-1$. When $\\Omega$ is smooth, bounded, pseudoconvex, and of finite type (the maximal order of contact of the boundary with complex analytic varieties is finite), the paper obtains the quantitative eigenvalue estimate $|\\lambda^{t,q}_k(\\Omega)-\\lambda^q_k(\\Omega)|\\le C\\,t\\,k\\,(\\lambda^q_k(\\Omega))^{2([1/(2\\alpha)]+1)}$, where $\\alpha\\in(0,1/2]$ is the order of subellipticity. It also proves the quantitative domain-perturbation bound $|\\lambda^{t,q}_k(\\Omega_j)-\\lambda^{t,q}_k(\\Omega)|\\le C_k\\,\\delta_j/t^{2n+3-1}$ for smooth bounded pseudoconvex domains $\\Omega_j$ whose normalized defining functions are uniformly bounded and $C^2$-close to that of $\\Omega$. The central mechanism is the min-max characterization of variational eigenvalues together with a transition operator between forms on the two domains, and Sobolev estimates for eigenforms supplied by subellipticity.","pith_inferences":["The same transition-operator scheme should extend to other non-coercive boundary value problems that admit subelliptic estimates, such as the tangential CR complex, yielding analogous $t$ and $\\delta$ rates.","The explicit $t$-rate suggests a practical recipe for numerical computation of $\\bar\\partial$-Neumann eigenvalues: to reach accuracy $\\varepsilon$, choose $t$ of order $\\varepsilon/(k(\\lambda^q_k)^p)$ with $p=2([1/(2\\alpha)]+1)$; the hidden constants in the Sobolev bounds depend on the domain and are not tracked, so the recipe is only asymptotic.","On model domains such as the unit ball, where explicit eigenvalues are known, one could test whether the powers of $t$ and $\\delta$ in the paper's bounds are optimal; the paper does not address optimality.","Because the proof of the eigenvalue convergence needs only a subelliptic estimate, the same quantitative result should hold uniformly for families of domains sharing a single subellipticity constant $\\alpha$, a uniformity statement not made in the paper."],"forward_implications":["For any bounded $C^2$ domain, every variational eigenvalue of the regularized problem converges to the corresponding $\\bar\\partial$-Neumann eigenvalue as $t\\to 0^+$, so the regularized spectrum can be made arbitrarily close to the true one.","On smooth bounded pseudoconvex finite-type domains, the $k$-th eigenvalue error is at most $C t k (\\lambda^q_k)^{2([1/(2\\alpha)]+1)}$; for fixed $k$ this is $O(t)$, and the bound degrades with both $k$ and the reciprocal of the subellipticity constant.","When the domain is perturbed by $\\delta_j$ in the $C^2$ norm, the regularized eigenvalues move by at most $C_k\\delta_j/t^{2n+3-1}$, giving quantitative control of domain dependence for the coercive problem.","Norm resolvent convergence holds for smoothly bounded strongly pseudoconvex domains, while on weakly pseudoconvex domains containing boundary complex varieties one cannot expect norm resolvent convergence because the regularized resolvent is compact but the $\\bar\\partial$-Neumann resolvent is not."],"supporting_citations":[{"why":"Introduces the elliptic regularization operator whose spectral stability is the subject; the added t-gradient term produces a coercive boundary value problem.","marker":"[KN65]"},{"why":"Supplies the subelliptic estimates on smooth bounded pseudoconvex domains of finite type that yield the Sobolev and C^l bounds on eigenforms used in Lemma 3.1.","marker":"[Ca87]"},{"why":"Provides Lemma 2.2, the transition-operator eigenvalue comparison, and the domain-perturbation construction that Theorem 4.3 adapts to the regularized operator.","marker":"[FZ19]"},{"why":"Gives the elliptic regularity estimate for the regularized operator (Proposition 3.5) whose constants Lemma 4.1 tracks explicitly as powers of t.","marker":"[S10]"},{"why":"Supplies the L^2 estimates for the ∂-operator that give the uniform lower bound on the ∂-Neumann form and the resolvent bounds used in Remark 1 and Theorem 3.3.","marker":"[H65]"},{"why":"Provides the definitions and standard facts about norm and strong resolvent convergence used to formulate and prove Theorem 3.3.","marker":"[RS80]"},{"why":"Defines the finite-type condition (bounded order of contact) that characterizes when the subelliptic estimates used in Theorem 1.1 apply.","marker":"[Dan82]"},{"why":"Sets up the ∂-Neumann problem, its domain spaces, and the variational eigenvalue framework on which the paper builds.","marker":"[CS99]"}],"fun_headline_variants":["Explicit rates for ∂-Neumann spectral convergence","Eigenvalue stability under elliptic regularization quantified","Quantitative domain-perturbation bounds for ∂-Neumann spectrum","Strong resolvent limits with t→0 rates for ∂-Neumann","Finite-type pseudoconvex domains: spectral closeness with rates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is the unproved reverse direction of the domain-perturbation estimate: the paper assumes, by symmetry, that swapping the two domains gives the same bound with the same constants, but the proof is left to the reader.","fun_headline_variants_meta":{"raw":{"variants":["Explicit rates for ∂-Neumann spectral convergence","Eigenvalue stability under elliptic regularization quantified","Quantitative domain-perturbation bounds for ∂-Neumann spectrum","Strong resolvent limits with t→0 rates for ∂-Neumann","Finite-type pseudoconvex domains: spectral closeness with rates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000188,"raw_usage":{"total_tokens":1306,"prompt_tokens":896,"completion_tokens":410,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":512,"completion_tokens_details":{"reasoning_tokens":326}},"tokens_in":512,"tokens_out":410,"duration_ms":4413,"temperature":1.0,"reasoning_tokens":326,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:20:04.254124+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the low eigenvalues of $\\square^t_q$ on a domain with explicitly known $\\bar\\partial$-Neumann spectrum, such as the unit ball in $\\mathbb{C}^n$, for a decreasing sequence of $t$; the quantitative part of Theorem 1.1 requires $|\\lambda^{t,q}_1-\\lambda^q_1|=O(t)$, so observing decay slower than a constant times $t$ would falsify it. For Theorem 1.2, build a family $\\Omega_j$ converging to $\\Omega$ in $C^2$ and check whether $|\\lambda^{t,q}_1(\\Omega_j)-\\lambda^{t,q}_1(\\Omega)|/\\delta_j$ stays bounded by $C/t^{2n+3-1}$ as $\\delta_j\\to 0$; unboundedness would disprove the bound.","supporting_citations":[],"review_version":1}