Spectral regularity with respect to dilations for a class of pseudodifferential operators

Horia D. Cornean; Radu Purice

REVIEW 1 major objections 4 minor 7 references

Spectral regularity with respect to dilations for a class of pseudodifferential operators

T0 review · 1 major / 4 minor · reviewed 2026-08-12 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read The paper proves that the spectral edges of a Weyl-quantized pseudodifferential operator move under dilation-type perturbations with rate (1+μ)/(2+μ), improving the universal square-root bound.

desk verdict The square-root Hausdorff bound is solid; the edge-regularity theorem is a genuine but under-supported claim, with the key estimate (4.18) unproved and the abstract overstating the result. read the letter →

arxiv 2411.14824 v1 pith:UUP2Z6OS submitted 2024-11-22 math.AP math-phmath.MP

Horia D. Cornean , Radu Purice This is my paper · ORCID

classification math.APmath-phmath.MP MSC 35S0547G3047A10

keywords pseudodifferentialoperatorsWeylquantizationspectralstabilitysymbolclassS^0_{00}dilationperturbationsedgesHausdorffdistance

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

When a smooth vector field F with all derivatives bounded is used to dilate the position variable of a real pseudodifferential symbol, x → x+δF(x), the spectrum of the corresponding Weyl-quantized operator is shown to stay within a square-root distance of the unperturbed spectrum. The paper's main new result is an edge regularity statement: if the second derivatives of F decay at infinity like a power 1+μ, then the top and bottom of the spectrum move with the faster rate |δ|^{(1+μ)/(2+μ)}, an exponent between 1/2 and 1. This removes the 'slow variation' hypothesis of the earlier paper, replacing it by a purely analytic decay condition on F. The argument works for all real symbols of the class $S^{0}$_{0,0} and yields explicit constants depending only on the symbol seminorms and F.

What carries the argument

The proof rests on two constructions. For the Hausdorff bound, a quasi-resolvent T(z;δ)=Σ_γ G_γ τ_{−z_γ}(K_0−z)^{-1}τ_{z_γ}G_γ is assembled from translations τ_{z_γ} at lattice points z_γ=δ^κ γ and a quadratic partition of unity G_γ = multiplication by g($δ^{{1−κ}}$F(x)−γ). A Newton–Leibniz expansion of the kernel difference shows the error is O(δ^κ)+O($δ^{{1−κ}}$), optimised at κ=1/2 to give √δ. For the edge estimate the paper changes tactic and works with quadratic forms, replacing the dilation x→x+δF(x) by the translation x→x+δF(u) and using a Gaussian weight identity to localise the difference. A two-zone cutoff (inner region with χ_θ and outer region with 1−χ_θ) separates the regime where the field is effectively constant from the regime where the second derivatives' decay can be traded against a growing δ-dependent weight; balancing the parameters θ and κ yields the exponent (1+μ)/(2+μ).

What would settle it

Take F(x)=(1+|x|^2)^{-μ/2}x so that ∂^2F decays like ⟨x⟩^{-(1+μ)}, choose a real symbol a in $S^{0}$_{0,0} whose Weyl quantization has a non-degenerate spectral edge, and compute E_+(δ)-E_+(0) numerically for δ→0; an observed scaling with exponent strictly smaller than (1+μ)/(2+μ) (meaning a slower approach to zero than the theorem allows) would contradict Theorem 2.3.

Watch

Extended reading notes

Core claim

The central discovery is Theorem 2.3: for a real symbol a of the class $S^{0}$_{0,0} and a smooth dilation field F with all derivatives bounded, if |∂_{x_j}∂_{x_k}F(x)| ≤ C⟨x⟩^{-(1+μ)} for some μ>0, then the spectral edges E_+(δ)=sup σ(Op^w(a(x+δF(x),ξ))) and E_-(δ)=inf σ(Op^w(a(x+δF(x),ξ))) satisfy |E_±(δ)−E_±(0)| ≤ C(a,F)|δ|^{(1+μ)/(2+μ)} for |δ|≤δ_0. Together with Theorem 2.1, which bounds the Hausdorff distance of the full spectra by C(a,F)√|δ|, this shows that the interior of the spectrum can move at the slower square-root scale while the edges enjoy a strictly better Hölder regularity controlled by the decay of F's second derivatives.

Load-bearing premise

The load-bearing premise is that the second derivatives of the perturbing field F decay at infinity at least as fast as the reciprocal of distance raised to the power 1+μ.

Editorial extensions

If this is right

The Hausdorff distance between σ(K_δ) and σ(K_0) is at most C(a,F)√|δ| for every real symbol in S^0_{0,0} and every F with globally bounded derivatives, and counterexamples show this scale can be attained by spectral gaps.
If the second derivatives of F decay like ⟨x⟩^{-(1+μ)}, then the supremum and infimum of the spectrum satisfy |E_±(δ)-E_±(0)| ≤ C(a,F)|δ|^{(1+μ)/(2+μ)}.
The exponent ν=(1+μ)/(2+μ) interpolates between 1/2 at μ→0 and 1 at μ→∞, so the edge motion ranges from barely better than square-root to almost Lipschitz.
The constants depend only on finitely many symbol seminorms and on the derivative bounds of F, so the estimates are uniform in the perturbation parameter δ.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

A natural test case is F(x)=(1+|x|^2)^{-μ/2}x, for which the decay condition is exactly satisfied; computing E_+(δ) numerically for a simple symbol would show whether the exponent (1+μ)/(2+μ) is sharp or merely an upper bound.
The same Gaussian-weight localisation could be applied to more general perturbations x↦x+δG(x) that are not pure dilations but share the decay of second derivatives, suggesting a wider class of 'asymptotically affine' perturbations with edge regularity.
The paper proves the improved rate for the global spectral edges only; whether internal gaps open or close with the same rate is left open by the argument, since the quadratic-form estimates are global rather than localised to a gap.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

Short version: Thm 2.1 is good, Thm 2.3 is the interesting new claim but its proof has a load-bearing gap in Prop 4.2, and the abstract oversells the edge behavior.

What's actually new: the symbol class a(x+δF(x),ξ) with general bounded-derivative F, removing the slow-variation assumption of [2]. The O(√δ) Hausdorff distance (Thm 2.1) is a clean, self-contained argument using quasi-resolvents and localization; I see no issue there. The edge exponent (1+μ)/(2+μ) under the decay of F'' is a genuinely new target.

Where it breaks: (4.18) asserts |I1| ≤ κ^{-2}θ^{1+μ} without proof. I1 contains (∇⊗∇)Fθ^⊥ after the first-order term is removed by symmetry of Wκ. On the support of 1−χθ, (1−χθ)∂²F is O(θ^{1+μ}), but differentiating the cutoff gives −θχ'(θz)∂F and −θ²χ''(θz)F, supported on |z|∼θ^{-1}, and these are O(θ) because F may grow linearly. No cancellation is shown. If they survive, the middle term in (4.12) becomes δ^{2−2ρ} rather than δ^{(2+μ)(1−ρ)}. For μ>1 the stated exponent is then not reachable by this argument; for μ<1 the argument can still produce a rate at least 2/3, which is stronger than the claim, so the theorem may survive in that range, but the paper's derivation of the exact exponent is wrong. The parameter line "ρ=(2+μ)(1−ρ) implies ρ=(1+μ)/(2+μ)" is also algebraically misleading as printed; likely a typo, but it should be corrected. Finally, Theorem 2.3 only bounds sup and inf of the spectrum, while the abstract and introduction talk about spectral edges more broadly, and 'behaves like' implies a two-sided statement when only an upper bound is proven.

Bottom line: the problem is meaningful and the strategy is serious. I would not desk-reject it, but it needs a substantive revision: prove or repair (4.18), weaken the statement if necessary, and align the abstract with what is actually shown.

Referee Report

1 major / 4 minor

Summary. The paper studies the spectral stability of Weyl pseudodifferential operators with symbols a(x + δF(x), ξ), where a ∈ S^0_{0,0}(R^d × R^d) is real and F ∈ C^∞_1(R^d; R^d) has bounded derivatives of positive order. Theorem 2.1 establishes that the Hausdorff distance between the spectra of the perturbed and unperturbed operators is O(√|δ|). Theorem 2.3 claims that if the second derivatives of F decay as ⟨x⟩^{-(1+μ)} for some μ>0, then the spectral edges E_±(δ) satisfy |E_±(δ) − E_±(0)| ≤ C(a,F)|δ|^{(1+μ)/(2+μ)} for small δ. The proof of Theorem 2.1 uses a quasiresolvent built from a lattice of translated resolvents and commutator estimates. The proof of Theorem 2.3 uses quadratic forms, a Gaussian weight W_κ, and two cutoffs (one on the kernel variable v, one on the field F). The central estimate is Proposition 4.2, which bounds the difference between the quadratic forms of the original and weighted operators by terms δ/θ, δκ^{-2}θ^{1+μ}, and δ^2κ^{-2}; the optimal choice of θ and κ then yields the claimed exponent.

Significance. If the result of Theorem 2.3 is correct, it gives a quantitative spectral edge regularity that goes beyond the general square-root Hausdorff bound and explicitly ties the edge rate to the spatial decay of the perturbation's curvature. This is a meaningful refinement of the authors' earlier work on slowly varying perturbations and is of interest to the spectral theory community. The proof of Theorem 2.1 is coherent and appears sound: the quasiresolvent construction, the use of a quadratic partition of unity, and the balancing of the two commutator estimates at κ=1/2 are natural and well executed. The paper is self-contained, reproducing a needed lemma from the authors' previous paper, and does not rely on any ad hoc or circular assumptions. However, the main new result, Theorem 2.3, rests on an unproved estimate for the outer-region integral I1 in Proposition 4.2. As written, that estimate is not justified under the stated hypotheses, so the claimed exponent is not established. This gap is load-bearing and requires a substantial repair or a revision of the theorem's statement.

major comments (1)

[4, Proposition 4.2, eq. (4.18)] The bound |I1[φ]| ≤ κ^{-2}θ^{1+μ} is not justified by the preceding text. In the definition of I1 (eq. (4.17)), the term involving (∇⊗∇)F^⊥_θ is understood pointwise, but F^⊥_θ = (1 − χ_θ)F is a product, so its Hessian contains the cutoff derivatives −2∇χ_θ ⊗ ∇F − (Δχ_θ)F in addition to (1−χ_θ)∇⊗∇F. These extra terms are supported on the annulus |z| ∼ θ^{-1}. Under the hypotheses, ∇F is bounded and F may grow linearly, so these terms are only O(θ), not O(θ^{1+μ}); specifically, F∇χ_θ is O(θ · θ^{-1}) = O(1) at the level of first derivatives and the Hessian terms are O(θ) due to the θ^2 factors from Δχ_θ combined with F ∼ θ^{-1}. No cancellation or integration by parts is shown to eliminate these contributions. If |I1| is only bounded by Cκ^{-2}θ, then the term δ|I1| becomes δ^{2−2ρ} after the parameter choices κ^2 = δ^ρ, θ = δ^{1−ρ}, and the optimization yields the rate δ^{2/3} rather than the claimed δ^{(1+μ)/(2+μ)} for 0 < μ < 1 (and also for μ > 1, where the claimed exponent exceeds 2/3). The proof must either provide a rigorous estimate for I1 that accounts for the cutoff derivatives or the statement of Theorem 2.3 must be weakened accordingly. The current text simply asserts (4.18) after defining I1 and I2, which is insufficient for a result of this precision.

minor comments (4)

[Section 4, parameter choices] The line 'ρ = (2+μ)(1−ρ) implies ρ = (1+μ)/(2+μ)' is algebraically wrong: the solution is ρ = (2+μ)/(3+μ). Since this correct value gives a larger exponent than the one stated (δ^{(2+μ)/(3+μ)} < δ^{(1+μ)/(2+μ)} for 0<δ<1), the claimed estimate still follows, but the derivation should be corrected.
[Abstract and Notation 1.2] The abstract says F has 'all its derivatives globally bounded', which could be misread to include F itself, but Notation 1.2 defines C^∞_1 as bounded derivatives of strictly positive order. This discrepancy should be resolved, e.g., by saying 'all derivatives of positive order' in the abstract as well.
[Throughout] The manuscript contains numerous typos and OCR-like artifacts, such as 'evidengtly' in the proof of Proposition 3.3, misaligned parentheses in (3.1), and inconsistent use of slashes in formulas. A careful proofreading pass is needed before publication.
[Section 4, proof of Lemma 4.3/4.4] In Lemma 4.3 and Lemma 4.4, the bound ν_{n,m}(a'_{δ,θ}) ≤ M F θ^{-1} sup_{0≤s≤1} ν_{n+1,m}(a[F]_s) uses the sup of |F_θ|. While this is correct, it would be clearer to explicitly state that sup|F_θ| ≤ Cθ^{-1} because F grows at most linearly on the support of χ_θ.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: Theorem 2.3 is derived from Weyl calculus and symbol seminorm estimates; the self-citations to [2] are either reproduced with proof or used only for motivation and notation.

full rationale

The central result, Theorem 2.3, is obtained by estimating quadratic-form differences between the perturbed kernel Kδ and the smoothed kernel Wκ[Kδ]. The edge-rate exponent (1+μ)/(2+μ) is not an input to any definition or fit: it emerges from the parameter balancing θ=δ^{1-ρ}, κ²=δρ at the end of Section 4, using the asserted bounds in (4.12) and (4.18). No spectral quantity is fitted to data and then renamed a prediction, and no theorem is imported solely by citation: Lemma 3.2 explicitly reproduces the proof of Lemma 2.4 from [2], and the weight function Wκ used in Section 4 is defined in the paper itself, with the identity (4.4) verified there. The citations to [2] and [4] provide context and motivation, not load-bearing justification for the main estimate. The skeptic's concern about the unproved bound |I1|≤κ^{-2}θ^{1+μ} in (4.18), and the apparent algebraic slip in the parameter equation, would be correctness or rigor issues, not circularity: even if the bound is unjustified, it is not an input to the theorem restated as the conclusion. The derivation chain is self-contained against the stated assumptions, so the circularity score is 0.

Assumptions & free parameters 0 free parameters · 7 assumptions · 0 invented entities

The paper introduces no free parameters or invented entities. It relies on standard Weyl calculus, Calderon-Vaillancourt boundedness, functional calculus for self-adjoint bounded operators, and explicit assumptions on the perturbation field F. Scaling parameters κ, θ, ρ in the proof are optimization choices, not additional hypotheses.

assumptions (7)

standard math Calderon-Vaillancourt theorem: Op(a) is bounded on L2 for a in S^0_{0,0} with operator norm controlled by a finite seminorm.
Invoked in (1.4) and in the boundedness criterion (1.8).
standard math Weyl quantization kernel representation and Schwartz kernel theorem for symbols in S^0_{0,0}.
Used throughout, e.g., equations (1.3)-(1.7) and Proposition 1.1.
standard math Functional calculus for self-adjoint bounded operators and unitarity of translations.
Used in the quasi-resolvent construction (3.4) and Remark 3.1.
standard math Existence of a smooth compactly supported partition of unity with finite overlap.
Used in the localization procedure before Lemma 3.2.
domain assumption F is smooth with all positive-order derivatives globally bounded.
Stated in 'The Problem' and used in Remark 1.3 and the localization estimates.
domain assumption Second derivatives of F decay as |∂xj∂xk F(x)| ≤ C <x>^{-(1+μ)}.
Assumption in Theorem 2.3, used in the outer-region estimates (4.17)-(4.18).
standard math Gaussian identity (4.4) for the weight Wκ.
Used in Proposition 4.1 to factor the quadratic form into a translation-conjugated operator.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Spectral regularity with respect to dilations for a class of pseudodifferential operators." pith.science (2026). https://pith.science/paper/UUP2Z6OS

@misc{pith2026241114824,
  author       = {Pith},
  title        = {Pith review of: Spectral regularity with respect to dilations for a class of pseudodifferential operators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UUP2Z6OS}},
  note         = {Machine review of arXiv:2411.14824}
}

abstract

We continue the study of the perturbation problem discussed in \cite{CP3} and get rid of the 'slow variation' assumption by considering symbols of the form $a\big(x+\delta\,F(x),\xi\big)$ with $a$ a real H\"{o}rmander symbol of class $S^0_{0,0}(\mathbb{R}^d\times\mathbb{R}^d)$ and $F$ a smooth function with all its derivatives globally bounded, with $|\delta|\leq1$. We prove that while the Hausdorff distance between the spectra of the Weyl quantization of the above symbols in a neighbourhood of $\delta=0$ is still of the order $\sqrt{|\delta|}$, the distance between their spectral edges behaves like $|\delta|^\nu$ with $\nu\in[1/2,1)$ depending on the rate of decay of the second derivatives of $F$ at infinity.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

7 extracted references · 7 canonical work pages

[2]

Journal of Spectral Theory

Cornean, H.D., Purice, R.: Sharp spectral stability for a class of singularly perturbed pseudo-diﬀerential operat ors. Journal of Spectral Theory. 13 (3), (2023), 1129 - 1144

work page 2023
[1]

Calder´ on, A.-P., Vaillancourt, R.: On the boundedness of pseudo-diﬀerential operators. J. Math. Soc. Japan 23, 374–378 (1971)

work page 1971
[3]

Gr¨ ochenig, K: Time-frequency analysis of Sj¨ ostrand’s class. Rev. Mat. Iberoam. 22(2), 703-724 (2006)

work page 2006
[4]

Journal of Spectral Theory

Gr¨ ochenig, K., Romero, J.L., Speckbacher, M.: Lipschi tz Continuity of Spectra of Pseudodiﬀerential Operators in a Weighted Sj¨ ostrand Class and Gabor Frame Bounds. Journal of Spectral Theory. 13 (2023), 805–839

work page 2023
[5]

H¨ ormander: The Analysis of Linear Partial Diﬀerential Operators III: Ps eudo-Diﬀerential Operators

L. H¨ ormander: The Analysis of Linear Partial Diﬀerential Operators III: Ps eudo-Diﬀerential Operators . Springer- Verlag Berlin Heidelberg, (2007)

work page 2007
[6]

Hermann, 1978

Laurent Schwartz: Th´ eorie des Distributions. Hermann, 1978

work page 1978
[7]

Taylor: Pseudodiﬀerential operators (PMS-34)

Michael E. Taylor: Pseudodiﬀerential operators (PMS-34) . Princeton Mathematical Series 97. Princeton Legacy Li- brary. Princeton: Princeton University Press, 1981. (H. D. Cornean) Department of Mathematical Sciences, Aalborg University Thomas Manns Vej 23, 9220 Aalborg, Denmark E-mail address: cornean@math.aau.dk (R. Purice) ”Simion Stoilow” Institute of...

work page 1981

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.

[1] [2]

Journal of Spectral Theory

Cornean, H.D., Purice, R.: Sharp spectral stability for a class of singularly perturbed pseudo-diﬀerential operat ors. Journal of Spectral Theory. 13 (3), (2023), 1129 - 1144

work page 2023

[2] [1]

Calder´ on, A.-P., Vaillancourt, R.: On the boundedness of pseudo-diﬀerential operators. J. Math. Soc. Japan 23, 374–378 (1971)

work page 1971

[3] [3]

Gr¨ ochenig, K: Time-frequency analysis of Sj¨ ostrand’s class. Rev. Mat. Iberoam. 22(2), 703-724 (2006)

work page 2006

[4] [4]

Journal of Spectral Theory

Gr¨ ochenig, K., Romero, J.L., Speckbacher, M.: Lipschi tz Continuity of Spectra of Pseudodiﬀerential Operators in a Weighted Sj¨ ostrand Class and Gabor Frame Bounds. Journal of Spectral Theory. 13 (2023), 805–839

work page 2023

[5] [5]

H¨ ormander: The Analysis of Linear Partial Diﬀerential Operators III: Ps eudo-Diﬀerential Operators

L. H¨ ormander: The Analysis of Linear Partial Diﬀerential Operators III: Ps eudo-Diﬀerential Operators . Springer- Verlag Berlin Heidelberg, (2007)

work page 2007

[6] [6]

Hermann, 1978

Laurent Schwartz: Th´ eorie des Distributions. Hermann, 1978

work page 1978

[7] [7]

Taylor: Pseudodiﬀerential operators (PMS-34)

Michael E. Taylor: Pseudodiﬀerential operators (PMS-34) . Princeton Mathematical Series 97. Princeton Legacy Li- brary. Princeton: Princeton University Press, 1981. (H. D. Cornean) Department of Mathematical Sciences, Aalborg University Thomas Manns Vej 23, 9220 Aalborg, Denmark E-mail address: cornean@math.aau.dk (R. Purice) ”Simion Stoilow” Institute of...

work page 1981