Harmonic Approximation and Resolvent Estimates for Semiclassical Non-Self-Adjoint Operators

Stepan Malkov

REVIEW 2 major objections 3 minor 1 cited by

Harmonic Approximation and Resolvent Estimates for Semiclassical Non-Self-Adjoint Operators

T0 review · 2 major / 3 minor · reviewed 2026-08-03 · deepseek-v4-flash

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

Pith's one-line read For a broad class of non-self-adjoint semiclassical operators, the paper proves an O(1/h) resolvent bound in the low-lying spectral region by averaging the real part of the symbol along the Hamiltonian flow of the imaginary part, localizing

desk verdict Genuinely new averaging method, but Theorem 1.1 is false as stated — the proof's λ′ shift moves the excluded set to h(Ω + p1(0,0)), and p = |ξ|² + x² + h is a clean counterexample. read the letter →

arxiv 2601.17643 v2 pith:QUB5BG5Q submitted 2026-01-25 math.SP math.APmath.CV

Stepan Malkov This is my paper · ORCID

classification math.SPmath.APmath.CV MSC 35P2035S0581Q2047A10

keywords non-self-adjointoperatorssemiclassicalanalysisresolventestimatespseudospectrumharmonicapproximationFBItransformBargmannspacemethodofaveraging

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

This paper proves a semiclassical resolvent estimate for non-self-adjoint, non-elliptic pseudodifferential operators whose principal symbol has nonnegative real part and a single critical point. The central idea is to average the real part of the symbol along the Hamiltonian flow of the imaginary part; under the assumption that this averaged symbol is elliptic away from the origin and behaves like a positive quadratic form near it, an exponential weight can be constructed that turns the operator into a weakly elliptic one. Working on the Bargmann space — the Gaussian-weighted holomorphic $L^2$ space — the author combines a local estimate near the origin, governed by the quadratic approximation of the symbol, with an exterior ellipticity estimate, and binds them with a scaling parameter $\epsilon = A h$ to obtain a resolvent bound of order $1/h$ for $\lambda$ of size at most $C h$ away from the spectrum of the quadratic approximation. A sympathetic reader would care because this relaxes earlier assumptions that the zero set of the real part be finite, and it yields concrete resolvent and spectral-localization statements for Schrödinger operators with complex potentials.

What carries the argument

The carrying object is the time-averaged symbol $\langle a \rangle_{b,T}(X) = \frac{1}{2T} \int_{-T}^{T} a(e^{t H_b}(X)) \, dt$. The proof constructs a compactly supported weight $G_\epsilon$ satisfying $H_{\operatorname{Im} p_0} G_\epsilon = \langle (\operatorname{Re} p_0)_\epsilon \rangle_{\operatorname{Im} p_0,T} - (\operatorname{Re} p_0)_\epsilon$, which, after an almost holomorphic extension of the symbol and a complex deformation $X + i \delta H G_\epsilon(X)$, makes the real part of the deformed symbol positive quadratically near the origin and weakly elliptic outside. On the Bargmann side, the quantization-multiplication formula converts this symbol ellipticity into norm estimates; the local region is handled by the exact spectrum of the quadratic operator $q^w(x,D)$ and an analytic rescaling, while the exterior r

What would settle it

For the explicit one-dimensional family $P = -h^2 \frac{d^2}{dx^2} + x^2 + i x^2$, which satisfies the hypotheses ($q(\xi,x) = \xi^2 + (1+i)x^2$ has a known spectrum), take a sequence $h = 2^{-k}$, choose $z_0$ outside $\operatorname{Spec}(q^w)$, and set $\lambda = h z_0$. Numerically compute $\| (P - \lambda)^{-1} \|_{L^2 \to L^2}$ for small $h$. If the norm grows faster than $C/h$ along this sequence, Theorem 1.1 is false; if it stays $O(1/h)$, the main claim is supported for this test case.

Watch

Extended reading notes

Core claim

The main result, Theorem 1.1, states that under assumptions (1.9), (1.13)-(1.15), (1.17), (1.19), and (1.20), for every open neighborhood $\Omega$ of the spectrum of the quadratic approximation $Q$ and every $C > 1$ there is $h_0 > 0$ such that for all $h \le h_0$ and $\lambda \in D(0, C h) \setminus h \Omega$, the resolvent $(P - \lambda)^{-1}$ exists and its $L^2$ to $L^2$ norm is $O(1/h)$. The discovery is that an $O(h)$ spectral localization and the associated resolvent bound follow not from ellipticity of $\operatorname{Re} p_0$ itself but from ellipticity of its time-average along the Hamilton flow of $\operatorname{Im} p_0$, a strictly weaker dynamical condition. The theorem extends the harmonic-approximation framework to a regime where the critical set is a s

Load-bearing premise

The proof rests on Lemma 3.2, cited to an external thesis, which asserts that the time average of $\operatorname{Re} p_0$ near the origin equals the averaged quadratic form up to $O(|X|^3)$; the entire weight construction and the four-region ellipticity estimates collapse if this cubic remainder bound fails.

Editorial extensions

If this is right

The spectrum of P in the low-lying region D(0, Ch) is confined, up to a shift by the subprincipal symbol p_1(0,0), to an O(h)-neighborhood of the spectrum of the quadratic approximation q^w(x,D).
The resolvent bound ||(P - lambda)^{-1}|| <= C/h holds for every lambda of size O(h) that stays away from that shifted lattice, giving a controlled inverse in the semiclassical limit.
For complex Schrödinger operators P = -h^2 Delta + V + iW satisfying V >= 0, V(0) = W(0) = 0, V^{-1}(0) cap (nabla W)^{-1}(0) = {0}, and a nondegenerate critical point of W at 0, the theorem yields the explicit resolvent bound (1.33) for small h.
Combined with the finite-dimensional spectral reduction announced by the author in a subsequent paper, the result yields a complete asymptotic expansion of the eigenvalue lattice in the O(h)-neighborhood of the origin.
The method relaxes the earlier finite-critical-set framework by allowing the zero set of the real part of the principal symbol to be any compact set, as long as the averaged symbol is elliptic away from the origin.

Reading between the lines

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

The averaging construction is local in phase space, so a partition of unity on the phase space should extend the single-point critical set to a finite set, yielding resolvent bounds with the spectrum near a finite union of quadratic eigenvalue lattices.
The dynamical condition (1.20) is a linear-gain ellipticity hypothesis; replacing it with a power-law gain would likely produce resolvent bounds of order h^{-alpha} for alpha > 1, interpolating between the elliptic and subelliptic regimes.
For the explicit Schrödinger family, the averaged symbols (1.31) and (1.32) are computable in closed form, offering a concrete way to test whether the O(1/h) constant is sharp and whether the subprincipal shift p_1(0,0) appears in numerical eigenvalue asymptotics.
The dependence of the proof on epsilon = Ah with A large suggests that a quantitative version of Theorem 1.1 could track how the resolvent constant degrades as the averaging time T varies, which would be testable in numerical experiments.

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

Colleague —

Two things to know. First, the averaging method is real: replacing the HPS finite-critical-set hypothesis {Re p0 = 0} finite by the dynamical condition on {Re p0 = 0, H_{Im p0} = 0} using averages along the Im p0 flow is a genuine relaxation, and the weight construction with the four-region ellipticity is substantial work. Second, the main theorem as stated is false, and the failure is exactly where the proof and statement don't line up.

In Section 7 the proof replaces eq^w by ep^w − h p1(0,0) and defines λ′ = λ + p1(0,0). What the argument actually establishes is: for the O(1) parameter λ avoiding Ω, one gets ∥(P − h(λ + p1(0,0)))^{-1}∥ ≤ C/h. In terms of the spectral parameter μ, that excludes μ ∈ h(Ω + p1(0,0)), not μ ∈ hΩ. The stress-test's sign on the shift is backwards, but its conclusion is right. Take p = |ξ|^2 + x^2 + h. All hypotheses hold; q = |ξ|^2 + x^2, Spec(Q) = odd integers; P = −h²Δ + x² + h has an exact eigenvalue 2h = h(1 + p1(0,0)). For any small Ω containing Spec(Q), 2h ∈ D(0,2h)\hΩ, and the resolvent does not exist. The theorem is overbroad. Fix: assume p1(0,0) = 0, or change the excluded set to h(Ω + p1(0,0)).

The other soft spots are smaller. Lemma 3.2, the O(|X|³) expansion of the averaged symbol, is cited to a PhD thesis with no statement of its hypotheses; it is load-bearing for Proposition 3.1 and should be proved or precisely quoted. The four-region argument is compressed, and the dense-range step for the adjoint is asserted rather than shown.

Credit where due: the dynamical critical-set condition and the explicit Schrödinger example class are new relative to [HPS10]/[HPS13]; the proof is a forward chain with no circularity and no fitted parameters. No code or formal proofs, so the credit rests on the argument, which is mostly sound modulo the statement bug.

Who this is for: specialists in non-self-adjoint semiclassical analysis. It deserves a serious referee — the method is salvageable and significant, and a referee can pin down the corrected statement. I would send it to review with instruction to fix the theorem before acceptance.

Referee Report

2 major / 3 minor

Summary. The paper claims a semiclassical resolvent estimate for a broad class of non-self-adjoint, non-elliptic h-pseudodifferential operators whose principal symbol has a single critical point and whose real part can be made elliptic after averaging along the Hamilton flow of the imaginary part. Theorem 1.1 asserts that for any neighborhood Ω of Spec(Q) and any C>1, the resolvent (P−λ)^{-1} exists and is O(1/h) for all low-lying λ∈D(0,Ch)∖hΩ. The proof combines four ingredients: a reduction to symbols in S(1), an averaging construction producing a weight function with improved ellipticity on a complex deformed phase space, a local quadratic resolvent estimate on the Bargmann side, and an exterior ellipticity estimate; these are glued in Section 7. The manuscript also gives a Schrödinger operator example satisfying the dynamical hypotheses.

Significance. If established in corrected form, the result would be a meaningful generalization of the HPS10/HPS13 framework: it allows non-elliptic principal symbols with a finite critical set and uses the imaginary-part flow to obtain hypoelliptic averaging. The proof is ambitious and technically rich, combining FBI/Bargmann transformations, complex IR-manifolds, and quadratic spectral theory; it is also free of fitted parameters and proceeds by explicit dynamical hypotheses. However, the central theorem as stated is false, and the proof actually establishes a shifted version of the statement. Moreover, a key averaging expansion is quoted from an unpublished Ph.D. thesis without stating its hypotheses. These issues are load-bearing for the paper's main claim, so the manuscript cannot be accepted in its present form.

major comments (2)

[§7, Theorem 1.1] Equation (7.19) proves the resolvent bound for p^w − hλ′ with λ′ := λ + p1(0,0), where the local analysis requires λ to avoid Spec(Q). For the spectral parameter L appearing in Theorem 1.1, one has L = hλ′, so the proof actually requires L/h − p1(0,0) ∉ Spec(Q), not merely L/h ∉ Ω. The mismatch is not negligible because L is O(h) and p1(0,0) is O(1). Concretely, take q=|x|²+|ξ|², p0=q, and p=q+hα with α∉Spec(Q)−Spec(Q). Then p1(0,0)=α, and P=q^w(x,hD)+hα is unitarily equivalent to h(q^w(x,D)+α), so h(E_j+α) are exact eigenvalues for every E_j∈Spec(Q). Choose Ω to be a union of small disjoint neighborhoods of Spec(Q) avoiding E_1+α. Then λ0=h(E_1+α) lies in D(0,Ch)∖hΩ for small h, yet the resolvent does not exist at λ0. Thus Theorem 1.1 is false as stated. The theorem must be restated with the excluded set shifted by p1(0,0), i.e. λ/h−p1(0,0)∉Ω, or the hypothesis p1(0,0)=0 must be added.
[§3, Lemma 3.2] Lemma 3.2 is the keystone of Proposition 3.1: it supplies the expansion ⟨Re p0⟩_{Im p0,T}(X)=eq(X)+O_T(|X|³) that is used in the near-origin region (3.26)–(3.27) and in the intermediate region. The proof is entirely deferred to the Ph.D. thesis [Sto22], and the manuscript neither states the precise result proved there nor verifies that the hypotheses of Theorem 1.1 imply the hypotheses of that result. This is not a routine citation: it is the bridge that makes the averaged real part positive definite near the origin. The authors should either prove Lemma 3.2 under the stated assumptions or provide the exact theorem statement from [Sto22] and a detailed verification of its hypotheses.

minor comments (3)

[§7] The notation λ and λ′ is confusing: the proof switches between the rescaled spectral parameter λ (of order 1, measured against Spec(Q)) and the actual spectral parameter hλ′ of the operator. This notational ambiguity is directly related to the mismatch in Theorem 1.1 and should be fixed with distinct symbols (e.g. μ for the rescaled parameter and λ for the operator spectral parameter).
[§2] The sentence 'Then, the resolvent estimate for P yields...' appears before the resolvent estimate for P has been proved; consider reordering so that the a priori estimate is introduced after the reduction to S(1) is justified.
[Proposition 2.2] In (2.13)–(2.14), the Gronwall argument is written somewhat elliptically: the sup bound with Ct<1 on both sides is valid only after a standard continuation argument. Providing the standard Gronwall inequality explicitly would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the resolvent estimate is derived forward from explicit dynamical hypotheses, with load-bearing prior results cited from external authors.

full rationale

The paper's derivation chain is a forward argument from the standing assumptions (1.9), (1.13)-(1.15), (1.17), (1.19), and (1.20) to the resolvent bound. The dynamical conditions (1.19) and (1.20) are stated as hypotheses, not consequences of the theorem, and the proof uses them to construct the averaging weight and ellipticity estimates. The only imported result that is load-bearing is Lemma 3.2, which is delegated to the external Ph.D. thesis [Sto22]; it is not a self-citation, and no argument in the paper reduces the theorem to an equivalent restatement of that lemma. The method of averaging, Bargmann-space norms, and quantization-multiplication formula are likewise imported from [HPS10], [HSS05], and [HZ25], all external works and none authored by the present author. There are no fitted parameters renamed as predictions, no uniqueness theorem from the authors' own prior work invoked to force the ansatz, and no definition that already contains the conclusion. The apparent mismatch between the excluded set hΩ in Theorem 1.1 and the shifted set h(Ω - p1(0,0)) appearing in Section 7 is a possible correctness or statement-proof consistency issue, not a circularity of the kind this analysis flags. Accordingly, the appropriate finding is no significant circularity.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

The central claim rests on explicit symbol hypotheses and on a chain of prior results (HPS, HSS, Stone thesis, HZ). No free parameters or new physical entities are introduced. The only nonstandard imported result is Lemma 3.2, which is not proved in the text and is load-bearing for the averaging construction.

assumptions (5)

standard math Weyl pseudodifferential calculus, order functions, and symbol classes S(m), plus FBI/Bargmann transform and quantization-multiplication formula (Appendix A, [HZ25]).
Used throughout Sections 2, 4–7 as the technical foundation; accepted without proof.
standard math Analytic Fredholm theory and discreteness of spectra for quadratic operators ([HPS09], [Hel84]).
Used in Section 5 to justify the a priori estimate for q^w(x,D) away from its eigenvalues.
standard math Lemma 3.2: ⟨Re p0⟩_{Im p0,T}(X)=eq(X)+O_T(|X|^3) near the origin, imported from [Sto22].
This expansion underpins the ellipticity lower bounds in Proposition 3.1; no proof is given in the paper.
standard math Lemma 5.2: local resolvent localization estimate, imported from [HSS05], Proposition 5.2.
Used to cut the quadratic resolvent estimate to a neighborhood of the origin; not proved in the text.
domain assumption Assumptions (1.9), (1.13)–(1.15), (1.17), (1.19), (1.20) on p0 and the averaged symbol.
These define the class of operators covered by Theorem 1.1; they are explicit hypotheses, not derived.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Harmonic Approximation and Resolvent Estimates for Semiclassical Non-Self-Adjoint Operators." pith.science (2026). https://pith.science/paper/QUB5BG5Q

@misc{pith2026260117643,
  author       = {Pith},
  title        = {Pith review of: Harmonic Approximation and Resolvent Estimates for Semiclassical Non-Self-Adjoint Operators},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QUB5BG5Q}},
  note         = {Machine review of arXiv:2601.17643}
}

abstract

We study resolvent estimates and bounds on the low lying spectrum for a broad class of non-self-adjoint non-elliptic $h$-pseudodifferential operators with critical points. Imposing dynamical conditions on the average of the real part of the principal symbol along the Hamilton flow of the imaginary part, we establish precise semiclassical resolvent estimates in an $O(h)$-neighborhood of the boundary of the semiclassical pseudospectrum, away from the eigenvalues of quantizations of the quadratic approximations of the principal symbols of the operators.

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Semiclassical Schr\"odinger operators with purely imaginary potential
math.AP 2026-07 accept novelty 6.0 of 10

The leftmost eigenvalues of -h²Δ+iV are asymptotically iE+h^σμ where μ are eigenvalues of model operators at the most degenerate critical points of V, with σ=2α/(α+2).

Reference graph

Works this paper leans on

5 extracted references · cited by 1 Pith paper

[1]

1, 171–194

[Bel19] Ben Bellis,Subelliptic resolvent estimates for non-self-adjoint semiclassical Schrödinger operators, Journal of Spectral Theory9(2019), no. 1, 171–194. [BH24] Carl M. Bender and Daniel W. Hook,PT-symmetric quantum mechanics, Reviews of Modern Physics96 (2024), no. 4, 045002. [DS99] Mouez Dimassi and Johannes Sjöstrand,Spectral asymptotics in the s...

2019
[28]

1, 61–66

[Sim09] Barry Simon,Schrödinger operators with purely discrete spectrum, Methods of Functional Analysis and Topology15(2009), no. 1, 61–66. [Sjö02] JohannesSjöstrand,Lectures on resonances, Lecturenotes, 2002, WrittenversionoflecturesgivenatGothen- burg University and Chalmers (Spring 2000–2002). Dated Palaiseau, May 31,

2009
[1985]

4, 801–846

[HPS09] Michael Hitrik and Karel Pravda-Starov,Spectra and semigroup smoothing for non-elliptic quadratic oper- ators, Mathematische Annalen344(2009), no. 4, 801–846. [HPS10] ,Semiclassical hypoelliptic estimates for non-selfadjoint operators with double characteristics, Com- munications in Partial Differential Equations35(2010), no. 6, 988–1028. [HPS13] ...

2009
[2019]

Grebenkov and Bernard Helffer,On spectral properties of the Bloch–Torrey operator in two dimen- sions, SIAM Journal on Mathematical Analysis50(2018), no

[GH18] Denis S. Grebenkov and Bernard Helffer,On spectral properties of the Bloch–Torrey operator in two dimen- sions, SIAM Journal on Mathematical Analysis50(2018), no. 1, 622–676. [Hel84] Bernard Helffer,Théorie spectrale pour des opérateurs globalement elliptiques, Astérisque, no. 112, Société mathématique de France,

2018
[2022]

138, American Mathematical Society, 2012

[Zwo12] Maciej Zworski,Semiclassical analysis, Graduate Studies in Mathematics, vol. 138, American Mathematical Society, 2012

2012

Pith tools

Reviewed August 3, 2026 · model on record in the stance chip above.

[1] [1]

1, 171–194

[Bel19] Ben Bellis,Subelliptic resolvent estimates for non-self-adjoint semiclassical Schrödinger operators, Journal of Spectral Theory9(2019), no. 1, 171–194. [BH24] Carl M. Bender and Daniel W. Hook,PT-symmetric quantum mechanics, Reviews of Modern Physics96 (2024), no. 4, 045002. [DS99] Mouez Dimassi and Johannes Sjöstrand,Spectral asymptotics in the s...

2019

[2] [28]

1, 61–66

[Sim09] Barry Simon,Schrödinger operators with purely discrete spectrum, Methods of Functional Analysis and Topology15(2009), no. 1, 61–66. [Sjö02] JohannesSjöstrand,Lectures on resonances, Lecturenotes, 2002, WrittenversionoflecturesgivenatGothen- burg University and Chalmers (Spring 2000–2002). Dated Palaiseau, May 31,

2009

[3] [1985]

4, 801–846

[HPS09] Michael Hitrik and Karel Pravda-Starov,Spectra and semigroup smoothing for non-elliptic quadratic oper- ators, Mathematische Annalen344(2009), no. 4, 801–846. [HPS10] ,Semiclassical hypoelliptic estimates for non-selfadjoint operators with double characteristics, Com- munications in Partial Differential Equations35(2010), no. 6, 988–1028. [HPS13] ...

2009

[4] [2019]

Grebenkov and Bernard Helffer,On spectral properties of the Bloch–Torrey operator in two dimen- sions, SIAM Journal on Mathematical Analysis50(2018), no

[GH18] Denis S. Grebenkov and Bernard Helffer,On spectral properties of the Bloch–Torrey operator in two dimen- sions, SIAM Journal on Mathematical Analysis50(2018), no. 1, 622–676. [Hel84] Bernard Helffer,Théorie spectrale pour des opérateurs globalement elliptiques, Astérisque, no. 112, Société mathématique de France,

2018

[5] [2022]

138, American Mathematical Society, 2012

[Zwo12] Maciej Zworski,Semiclassical analysis, Graduate Studies in Mathematics, vol. 138, American Mathematical Society, 2012

2012