Eigenvalues of the Magnetic Neumann Laplacian on Domains with Peaks

Noah Koerner

REVIEW 1 major objections 6 minor 14 references

Eigenvalues of the Magnetic Neumann Laplacian on Domains with Peaks

T0 review · 1 major / 6 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read For cusp-shaped boundaries, magnetic Laplacian eigenvalues grow as λ^{2/(q+1)}, with sharpness q setting the rate.

desk verdict New asymptotic law for magnetic Neumann Laplacians on outward power-law peaks; the proof is solid, the exact-geometry assumption is the main limitation, and it deserves expert peer review. read the letter →

arxiv 2608.09469 v1 pith:CXYPMRZB submitted 2026-08-10 math.SP math.AP

Noah Koerner This is my paper · ORCID

classification math.SPmath.AP MSC 35P1535P20

keywords magneticNeumannLaplacianoutwardpeakeigenvalueasymptoticssharpnessorderone-dimensionalmodeloperatortype-IIsuperconductivitycurvilinearpolygonspectraltheory

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

The paper proves that on a bounded planar domain whose boundary has an outward cusp of power-law shape y ~ a x^q, the n-th eigenvalue of the magnetic Neumann Laplacian grows like $λ^{{2/(q+1)}}$ as the magnetic-field parameter λ tends to infinity, with prefactor (a_+-a_-)^{2/(q+1)} E_n(T) and remainder O($λ^{{2κ}}$). This matters because smooth boundaries and cornered polygons both give linear growth in λ; the cusp therefore sets the entire low-lying spectrum. The sharper the peak (larger q), the slower the divergence, matching the known trend that stronger singularities lower magnetic eigenvalues. The proof straightens the cusp by a coordinate change and shows that a one-dimensional model operator T with potential $s^{{2q}}$/12 + q(q-2)/(4s²) carries the leading order.

What carries the argument

The carrying object is the coordinate change Φ(s,t) = (s, s^q t), which maps the rectangle I×(a_-,a_+) onto the peak section V_I = {a_- x^q < y < a_+ x^q}. It turns the magnetic form into an expression on the rectangle whose leading part, after projection onto t-independent functions, is the one-dimensional model T_λ = -d²/ds² + ((a_+-a_-)^2/12) λ² $s^{{2q}}$ + q(q-2)/(4s²). The proof partitions the cusp into three scales: (0,$λ^{{-κ}}$), ($b_1λ^{{-κ}}$, $b_2λ^{{-1/(2q)}}$), and ($b_3λ^{{-1/(2q)}}$, b). The innermost section supplies eigenvalues of order $λ^{{2/(q+1)}}$; the two outer sections are lower-bounded by operators with faster growth. Min-Max comparison (Lemma 2.1) and the IMS formula transfer these bounds to the full domain, and optimizing the cutoff parameter κ yields the stated remainder.

What would settle it

Take the innermost section I=(0,$λ^{{-κ}}$) and compute E_1(Q^I_λ) numerically for increasing λ at fixed q, a_-, a_+ (for instance q=2). If E_1(Q^I_λ)/$λ^{{2/(q+1)}}$ does not converge to (a_+-a_-)^{2/(q+1)} E_1(T) with deviations no larger than O($λ^{{2κ}}$), the theorem's leading term is false. Alternatively, perturb the upper side to y = a_+ x^q + ε $x^{{q+δ}}$ and check whether the first eigenvalue changes at leading order; if it does, the exact-form hypothesis is indispensable.

Watch

Extended reading notes

Core claim

The central claim is that the low-lying spectrum of the magnetic Neumann Laplacian on a domain with an outward peak is asymptotically governed by the local power-law cusp alone. For each fixed n, E_n(N^Ω_λ) = $λ^{{2/(q+1)}}$ (a_+-a_-)^{2/(q+1)} E_n(T) + O($λ^{{2κ}}$), where κ = (2q+3)/((q+1)(2+3q)) and T is the Friedrichs extension of -d²/ds² + $s^{{2q}}$/12 + q(q-2)/(4s²) on L²(0,∞). The exponent 2/(q+1) is less than 1 for q>1, so the eigenvalues diverge slower than the linear-in-λ growth known for smooth domains and slower than the corner-driven growth known for curvilinear polygons. The result holds for all n simultaneously, meaning the cusp controls the whole bottom of the spectrum, not just the ground state.

Load-bearing premise

The load-bearing premise is that, in a fixed neighbourhood of the origin, the domain is exactly the straight power-law wedge V_b = {a_- x^q < y < a_+ x^q}; if the boundary only approximates that shape, or is perturbed by any subleading term, the theorem's constant and even the exponent are not established by the proof.

Editorial extensions

If this is right

Every fixed eigenvalue E_n(N^Ω_λ) has the same leading power λ^{2/(q+1)}; the peak determines the entire low-lying spectrum, not only the ground state.
A sharper peak (larger q) gives a smaller exponent 2/(q+1), so the divergence is slower and the model operator T fixes the corresponding constant for each n.
The section (0,λ^{-κ}) is the only region contributing at leading order; 'near the peak' and 'away from the peak' regions contribute only to the remainder.
The remainder bound O(λ^{2κ}) with κ=(2q+3)/((q+1)(2+3q)) quantifies how quickly the asymptotic law appears and is the price paid for the stronger singularity.

Reading between the lines

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

If the cusp is only asymptotically a power law, the same exponent might survive but the prefactor (a_+-a_-)^{2/(q+1)}E_n(T) should be checked against a subleading term such as ε x^{q+δ}; the paper's exact-form assumption does not cover this, and a dependence on ε at leading order would be a natural test.
The same straightening procedure should transfer to three-dimensional peaks or conical tips, producing a model operator with an angular coordinate and a scaling exponent set by the tip's power; this is a direct extension not stated in the paper.
In the superconductivity reading, cusped samples would be predicted to nucleate at lower field strength than smooth or cornered ones, because slower eigenvalue growth lowers the energy of the first magnetic state; this is a physical consequence the paper leaves implicit.
The model eigenvalues E_n(T) depend on q through s^{2q}/12 + q(q-2)/(4s²); tabulating them numerically for a range of q would produce a benchmark curve that direct finite-element computations on cusped domains could confirm or refute.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

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

The thing to know: this paper proves a genuinely new eigenvalue asymptotic for the magnetic Neumann Laplacian on planar domains with outward power-law peaks. The main term scales like λ^{2/(q+1)} instead of the λ order seen for smooth domains and corners, and the sharpness q enters the exponent. The model operator T is derived by unitary changes and projection, not fitted, and the upper and lower bounds match after choosing the IMS cutoff κ optimally. I found no circular reasoning or invented parameters.

What’s good: the structure is clear. The proof splits the peak into near, intermediate, and away regions; the near region gives the leading term and the other two regions are shown to be higher order. The one-dimensional model operator is natural and its properties (positivity, simplicity, comparison with truncated versions) are handled cleanly. The exposition is honest about where the result lives: it extends the Bonnaillie-Noël–Dauge program to zero-angle cusps.

The soft spots are real but not fatal. The main theorem only covers domains that coincide exactly with a straight-sided power-law peak in a fixed neighbourhood of the origin (Ω∩(−b,b)² = V_b, Definition 4.1). That is a strong geometric hypothesis: no boundary perturbations, no merely asymptotic power-law cusps. The paper does not address whether the constant or the exponent are stable under perturbations, and it does not claim to. Also, several steps invoke external results — [3, Thm 2.1], [5, App. B], [12, Lem. 8] — without stating the exact hypotheses. I don’t doubt these uses, but an expert referee should check them, especially [12] because the density claim is load-bearing for the lower bound.

One more check: the choice of κ in Theorem 1.1 balances the two remainders 2κ and (2q+3)/(q+1)−3qκ, giving κ=(2q+3)/((q+1)(2+3q)), which is inside the allowed interval for q>1. I verified that. The positivity of the model operator relies on the Hardy inequality and holds for q>1 even though the potential q(q−2)/(4s²) is negative for 1

Who this is for: researchers in spectral asymptotics for magnetic Schrödinger operators, especially those working on Ginzburg–Landau linearizations. It is a serious, competently executed proof. I would send it to an expert referee rather than desk reject.

Recommendation: engage with it; the referee should focus on the external citations and the long estimates in Section 3. The core claim looks solid.

Referee Report

1 major / 6 minor

Summary. The manuscript studies the magnetic Neumann Laplacian with constant magnetic field strength lambda on bounded planar domains having an outward power-law peak V_b = {a_- x^q < y < a_+ x^q}. Theorem 1.1 claims that for the n-th eigenvalue, E_n(N_lambda^Omega) = lambda^{2/(q+1)} (a_+ - a_-)^{2/(q+1)} E_n(T) + O(lambda^{2 kappa}) with kappa = (2q+3)/((q+1)(2+3q)), where T is the one-dimensional operator -h'' + h(s^{2q}/12 + q(q-2)/(4s^2)). The proof derives a one-dimensional model operator by explicit changes of variables and gauge transformations, establishes matching upper and lower bounds (Propositions 4.2 and 4.3) after splitting the peak into three regions, and optimizes the remainder exponent. The paper is clearly written and the proof strategy is coherent; however, I find a sign error in the s^{-2} coefficient of the model operator that changes the limiting constant E_n(T) and therefore the main theorem as stated.

Significance. If corrected, this would be a genuine extension of the magnetic Neumann Laplacian asymptotics from smooth boundaries and corners to zero-angle cusps. The leading-order exponent 2/(q+1), the explicit dependence on a_+ - a_-, and the optimized remainder are concrete and falsifiable predictions. A particular strength is that the model operator is derived through unitary transformations and operator comparisons rather than fitted, and the upper and lower bounds match at the chosen kappa. The error found below affects the limiting constant, not the overall method, so the result is likely repairable by a systematic sign correction.

major comments (1)

[Definition 2.3 and Lemma 3.4] The coefficient q(q-2)/(4s^2) in the model operator is incorrect; the calculation in Lemma 3.4 yields q(q+2)/(4s^2). For v = g(s) in the form ell_I^lambda of Lemma 3.3, the cross term coming from -q/(2s) g is -(q/(2s))(|g|^2)', and after integration by parts this contributes +q/(2s^2)|g|^2. Hence the total coefficient of s^{-2}|g|^2 is q^2/4 + q/2 = q(q+2)/4. For example, when q=2 and a_- = -a_+ = 1, the transformed form is integral |g'|^2 + (lambda^2 s^4/3 + 2/s^2)|g|^2 ds, not integral |g'|^2 + (lambda^2 s^4/3)|g|^2 ds. This error propagates into Corollary 3.5, the lower bound in Lemma 3.6 and Corollary 3.7, and the upper and lower bounds in Propositions 4.2 and 4.3, because all these arguments compare with the wrong operator T. The proof structure can likely be repaired by replacing q(q-2) with q(q+2) everywhere, but as written the constant E_n(T) in Theorem 1.1 is not the limiting constant for the actual operator.

minor comments (6)

[Definition 4.1 / Theorem 1.1] The theorem is proved only for domains that coincide exactly with V_b in a neighbourhood of the peak. The proof uses this exact equality in the IMS partition, in the construction of the tubular neighbourhoods, and in the inclusion V_I subset Omega; merely asymptotic power-law cusps or small perturbations of the sides are not covered. This is a scope limitation rather than an inconsistency, but it should be stated explicitly in the introduction.
[Lemmas 3.11, 3.12 and Proposition 4.3] Please state the precise hypotheses of the external results invoked: [5, Appendix B] for the unitary equivalence in curved strips, [12, Lemma 8] for the density of H^1_I(V_I), and [3, Theorem 2.1] for the lower bound on the cut-off curvilinear polygon, so that the reader can verify that they apply to the present peak geometries.
[Lemma 2.2] In the IMS formula, the term ||grad chi_j f||^2 should be written as integral |grad chi_j|^2 |f|^2 dx, or as || |grad chi_j| f ||^2, to avoid confusion with a derivative applied to the product f.
[Section 3 and Theorem 1.1] The symbol kappa is used both for the interval-endpoint exponent in Lemma 2.9 and Section 3 and for the optimized remainder exponent in Theorem 1.1; using distinct letters would improve readability.
[Lemma 3.8] The proof speaks of an eigenfunction of Q_I^lambda but then computes with the unitarily equivalent form ell_I^lambda; please make the notation consistent.
[References] Reference [12] lists an author as 'F. Sk'; this is presumably a typesetting corruption of 'F. Šk' or a similar name and should be corrected.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the model operator T is derived by unitary transformation and a matching two-sided estimate, not fitted or imported by self-citation.

full rationale

The derivation of the main term is self-contained in the sense relevant to circularity. Lemma 3.3 is an exact unitary change of variables on the peak section. Lemma 3.4 shows that the one-dimensional operator T_I^λ is obtained by restricting ℓ_I^λ to t-independent test functions, which is a legitimate variational upper bound via Lemma 2.1, not a fit: the coefficients (a_+-a_-)^2/12 and q(q-2)/4 come from explicit integration over the transverse interval, not from matching eigenvalues. The converse is proved in Lemma 3.6, where the projection P0 onto t-independent functions gives E_n(L_I^λ) ≥ E_n(T_I^λ)(1-o(1)), so the model operator is shown to control the true eigenvalues from both sides; no fitted parameter is renamed as a prediction. Proposition 2.4 derives λ^{2/(q+1)}(a_+-a_-)^{2/(q+1)} E_n(T) by unitary rescaling from T_λ, and the final remainder exponent is obtained by optimizing κ in Propositions 4.2-4.3. The paper cites [3], [5], [9], and [12] for standard IMS, tubular-neighbourhood, density, and corner estimates; none are self-citations of the author and none carry the leading-order claim. The exact equality Ω∩(-b,b)^2 = V_b in Definition 4.1 and Theorem 1.1 is a hypothesis restricting the theorem to the exact power-law peak class; this is a scope limitation, not a circular step. Therefore no circularity is present.

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

No fitted parameters appear; the geometry parameters q, a_+, a_- are inputs. The proof relies on standard functional analysis, three cited external theorems, and the exact geometric model of the peak. None of these axioms contains the target asymptotic, so the circularity burden is low.

assumptions (5)

standard math One-dimensional Hardy inequality (2.1)
Used throughout Sections 2 and 3 to prove positivity of the model operator T_lambda and to obtain lower bounds for truncated operators.
standard math External lower bound for curvilinear polygons from [3, Theorem 2.1]
Invoked in Proposition 4.3 to bound the away-from-the-peak region by a linear term c_1 lambda. The statement is not reproduced in the text.
standard math Unitary tubular-neighbourhood description from [5, Appendix B]
Used in Lemma 3.11 to transform the magnetic Laplacian near the curved sides of the peak into a model on a rectangle.
standard math Density result [12, Lemma 8] for H^1_I(V_I)
Used in Proposition 4.3 to justify that products of cutoff functions with H^1 functions lie in the domain of the mixed-boundary operator.
domain assumption Exact geometric form of the peak: Omega intersect (-b,b)^2 = V_b
Definition 4.1 postulates that the boundary in a fixed neighbourhood of the origin is exactly a power-law cusp with straight sides a_- x^q and a_+ x^q. The entire proof depends on this exact profile.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Eigenvalues of the Magnetic Neumann Laplacian on Domains with Peaks." pith.science (2026). https://pith.science/paper/CXYPMRZB

@misc{pith2026260809469,
  author       = {Pith},
  title        = {Pith review of: Eigenvalues of the Magnetic Neumann Laplacian on Domains with Peaks},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/CXYPMRZB}},
  note         = {Machine review of arXiv:2608.09469}
}

abstract

We consider the magnetic Neumann Laplacian on bounded domains in $\mathbb{R}^2$ with outward peaks. The operator is associated with a large magnetic field depending on a parameter $\lambda$, and we investigate the behaviour of the eigenvalues as $\lambda$ tends to $+\infty$. We show that their asymptotic expansion is influenced by the sharpness $q$ and geometry of the peak, and that its main term is of order $\lambda^{\frac{2}{q+1}}$. This is an extension of previous works on magnetic Neumann Laplacians in smooth domains and domains with corners.

Figures

Figures reproduced from arXiv: 2608.09469 by the authors.

**Figure 1.1.** Visualization of a peak and a curvilinear polygon with an outward peak angles α ∈ (0, π). So assuming that the curvilinear polygon with M corners with angles αj ∈ (0, 2π)\{π} has at least one convex corner, one has E := min i=1,...,M E1(N Sαi 1 ) < Θ0 and the asymptotic expansion of the first eigenvalue of the magnetic Neumann Laplacian on the curvilinear polygon is of the form E1(N Ω λ ) = Eλ + O(λ 1 2 ) as λ → +∞.… view at source ↗

**Figure 3.1.** Partition of V(0,b) Definition 3.2. Let U ⊆ R 2 be an open set and I ⊆ R an open interval and set H1 I (U) := {f ∈ H1 (U) | ∃ c0, c1 ∈ I, c0 < c1 : f(x, y) = 0, ∀ (x, y) ∈ U with x /∈ [c0, c1]} We define the quadratic form of the magnetic Laplacian on VI with mixed Dirichlet and Neumann boundary conditions as q I λ (f) = n VI λ (f), D(q I λ ) = the closure of H1 I (VI ) in H1 (VI ). As usual, QI λ refers to the self… view at source ↗

**Figure 3.2.** S ± λ , the one-sided tubular neighborhoods of Γ± between Πλ and the one-sided tubular neighourhoods S ± λ := ϕ±(Πλ) of sections of Γ±. A visualisation can be found on [PITH_FULL_IMAGE:figures/full_fig_p017_3_2.png] view at source ↗

Figures from the paper (1 more)

**Figure 3.3.** Figure 3.3: Visualizations of U ± λ and S ± λ |∂xχ ± λ | 2 ≤ |2(a+ − a−) −1 b −q 1 √ λa±qxq−1 ∥χ ′ ∥∞| 2 ≤ λ4(a+ − a−) −2 b −2q 1 a 2 maxq 2 b 2(q−1)∥χ ′ ∥ 2 ∞, |∂yχ ± λ | 2 ≤ |2(a+ − a−) −1 b −q 1 √ λ∥χ ′ ∥∞| 2 ≤ λ4(a+ − a−) −2 b −2q 1 ∥χ ′ ∥ 2 ∞. If we can show that χ ± λ f ∈ …

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

14 extracted references · 14 canonical work pages

[12]

Laplacian eigenvalues for large negative Robin parameters on domains with outward peaks

K. Pankrashkin, F. Sk, and M. Vogel. “Laplacian eigenvalues for large negative Robin parameters on domains with outward peaks”. In:Nonlinear Differential Equations and Applications NoDEA 32.129 (2025)

work page 2025
[1]

Onset of superconductivity in decreasing fields for general domains

A. Bernoff and P. Sternberg. “Onset of superconductivity in decreasing fields for general domains”. In:Journal of Mathematical Physics39.3 (1998), pp. 1272–1284

work page 1998
[2]

On the fundamental state energy for a Schr¨ odinger operator with magnetic field in domains with corners

V. Bonnailie-No¨ el. “On the fundamental state energy for a Schr¨ odinger operator with magnetic field in domains with corners”. In:Asymptotic Analysis41 (2005), pp. 215–258

work page 2005
[3]

Asymptotics for the low-lying eigenstates of the Schr¨ odinger operator with magnetic field near corners

V. Bonnaillie-No¨ el and M. Dauge. “Asymptotics for the low-lying eigenstates of the Schr¨ odinger operator with magnetic field near corners”. In:Annales Henri Poincar´ e7 (2006), pp. 899–931

work page 2006
[4]

Edmunds and W.D

D.E. Edmunds and W.D. Evans.Spectral theory and differential operators. Oxford: Clarendon Press, 1987

work page 1987
[5]

Accurate eigenvalue asymptotics for the magnetic Neumann Lapla- cian

S. Fournais and B. Helffer. “Accurate eigenvalue asymptotics for the magnetic Neumann Lapla- cian”. In:Annales de l’Institut Fourier56.1 (2006), pp. 1–67

work page 2006
[6]

Fournais and B

S. Fournais and B. Helffer.Spectral Methods in Surface Superconductivity. Birkhaeuser Boston, MA, 2010

work page 2010
[7]

Semiclassical analysis for the ground state energy of a Schr¨ odinger operator with magnetic wells

B. Helffer and A. Mohamed. “Semiclassical analysis for the ground state energy of a Schr¨ odinger operator with magnetic wells”. In:Journal of Functional Analysis138.1 (1996), pp. 40–81

work page 1996

Show all 14 references

[8]

Magnetic bottles in connection with superconductivity

B. Helffer and A. Morame. “Magnetic bottles in connection with superconductivity”. In:Journal of Functional Analysis185.2 (2001), pp. 604–680

2001
[9]

Bound states for the magnetic Neumann Laplacian in planar sectors

A. Kachmar and M. Sundqvist. “Bound states for the magnetic Neumann Laplacian in planar sectors”. In:arXiv2607.12600 (2026), https://arxiv.org/abs/2607.12600

2026 arXiv
[10]

Gauge invariant eigenvalue problems inR 2 andR 2 +

K. Lu and X. Pan. “Gauge invariant eigenvalue problems inR 2 andR 2 +”. In:Transactions of the American Mathematical Society352.3 (1999), pp. 1247–1276

1999
[11]

Estimates of the upper critical field for the Ginzburg–Landau equations of superconductivity

K. Lu and X.B. Pan. “Estimates of the upper critical field for the Ginzburg–Landau equations of superconductivity”. In:Physica D: Nonlinear Phenomena127.1 (1999), pp. 73–104

1999
[13]

Raymond.Bound States of the Magnetic Schr¨ odinger Operator

N. Raymond.Bound States of the Magnetic Schr¨ odinger Operator. European Mathematical So- ciety, 2017

2017
[14]

Weidmann.Spectral Theory of Ordinary Differential Operators

J. Weidmann.Spectral Theory of Ordinary Differential Operators. Springer Berlin, Heidelberg, 1987

1987

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.

[1] [12]

Laplacian eigenvalues for large negative Robin parameters on domains with outward peaks

K. Pankrashkin, F. Sk, and M. Vogel. “Laplacian eigenvalues for large negative Robin parameters on domains with outward peaks”. In:Nonlinear Differential Equations and Applications NoDEA 32.129 (2025)

work page 2025

[2] [1]

Onset of superconductivity in decreasing fields for general domains

A. Bernoff and P. Sternberg. “Onset of superconductivity in decreasing fields for general domains”. In:Journal of Mathematical Physics39.3 (1998), pp. 1272–1284

work page 1998

[3] [2]

On the fundamental state energy for a Schr¨ odinger operator with magnetic field in domains with corners

V. Bonnailie-No¨ el. “On the fundamental state energy for a Schr¨ odinger operator with magnetic field in domains with corners”. In:Asymptotic Analysis41 (2005), pp. 215–258

work page 2005

[4] [3]

Asymptotics for the low-lying eigenstates of the Schr¨ odinger operator with magnetic field near corners

V. Bonnaillie-No¨ el and M. Dauge. “Asymptotics for the low-lying eigenstates of the Schr¨ odinger operator with magnetic field near corners”. In:Annales Henri Poincar´ e7 (2006), pp. 899–931

work page 2006

[5] [4]

Edmunds and W.D

D.E. Edmunds and W.D. Evans.Spectral theory and differential operators. Oxford: Clarendon Press, 1987

work page 1987

[6] [5]

Accurate eigenvalue asymptotics for the magnetic Neumann Lapla- cian

S. Fournais and B. Helffer. “Accurate eigenvalue asymptotics for the magnetic Neumann Lapla- cian”. In:Annales de l’Institut Fourier56.1 (2006), pp. 1–67

work page 2006

[7] [6]

Fournais and B

S. Fournais and B. Helffer.Spectral Methods in Surface Superconductivity. Birkhaeuser Boston, MA, 2010

work page 2010

[8] [7]

Semiclassical analysis for the ground state energy of a Schr¨ odinger operator with magnetic wells

B. Helffer and A. Mohamed. “Semiclassical analysis for the ground state energy of a Schr¨ odinger operator with magnetic wells”. In:Journal of Functional Analysis138.1 (1996), pp. 40–81

work page 1996