Continuum limit of discretized matrix-valued Fourier multipliers

Mat\v{e}j Tu\v{s}ek; Ruben Karapetyan

REVIEW 2 major objections 4 minor 27 references

Continuum limit of discretized matrix-valued Fourier multipliers

T0 review · 2 major / 4 minor · reviewed 2026-08-01 · deepseek-v4-flash

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

Pith's one-line read This paper proves uniform norm resolvent convergence, with explicit rates, for discrete lattice approximations to a wide class of matrix-valued Fourier multipliers, including Dirac and bilayer graphene Hamiltonians.

desk verdict Main result is solid and a genuine extension; Theorem 3.10 has a real gap—its proof uses the very assumption it claims to drop. read the letter →

arxiv 2607.21177 v1 pith:DVBGPL4W submitted 2026-07-23 math-ph math.MPmath.SP

Ruben Karapetyan , Matěj Tušek This is my paper · ORCID

classification math-phmath.MPmath.SP MSC 47A5847B2581Q10

keywords normresolventconvergenceFouriermultipliersblockoperatormatricesDiracbilayergraphenefermiondoublingcorrectiontermlatticeapproximation

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 establishes uniform norm resolvent convergence for a wide class of matrix-valued Fourier multipliers when approximated on a lattice. The operators are self-adjoint block matrices of the form [[G11, G12], [G12*, −G11]]; the class includes Dirac operators in dimensions 1–3 and the effective Hamiltonian of bilayer graphene. Two discretization schemes are studied: a nonlocal one derived from the square root of the standard second difference, which converges in norm without any correction when the symbol is even in each momentum variable, and the symmetric-difference scheme, which converges in norm only after adding a correction term based on (−hΔ_h)^γ; without it, convergence fails in norm and holds only strongly. For symbols growing like |ξ|^(2γ) with derivatives growing at most like |ξ|^β, the error is of order h^{min{2, 2γ−max{β,0}−1}}, so the rates are explicit — O(h) for Dirac operators and O(h^2) for bilayer graphene.

What carries the argument

The argument turns on the block-diagonal form of G0^2, which follows from assuming the off-diagonal block G12 is normal and commutes with G11. This yields the resolvent bound (G0^2+1)^{-1} ≤ C(1+|ξ|^{2γ})^{-1}, and Taylor expansion controls the pointwise difference between each discrete symbol and the continuous one. For the symmetric-difference scheme, a correction term f_h(ξ) = (h∑ 4h^{-2} sin^2(hξ_j/2))^γ restores the lower bound at the Brillouin-zone boundary; without it, the pointwise resolvent difference stays O(1) there. The identification operators J_h, K_h built from biorthogonal generating functions turn the pointwise estimates into operator-norm bounds, with the neighboring-zone t

What would settle it

Compute the norm in (27) for the bilayer graphene symbol (γ=2, β=1) at h = 2^{-k}, k=1,...,8, using the sinc-function identification operators described in Example 1; Theorem 3.7 predicts a decay exponent of 2. If the fitted exponent is clearly below 2, the central bound is false.

Watch

Extended reading notes

Core claim

The central discovery is that, when the off-diagonal block G12 of the symbol is normal and commutes with the diagonal block G11, the square of the full matrix symbol becomes block-diagonal, yielding a uniform pointwise resolvent bound with decay |ξ|^(−2γ). The paper proves that two natural discrete approximations — a nonlocal square-root-of-Laplacian scheme and a symmetric-difference scheme with a specific correction term — converge in norm to the continuum resolvent with error O(h^{min{2,2γ−max{β,0}−1}}). It also proves that the uncorrected symmetric-difference scheme does not converge in norm, with the resolvent difference remaining O(1) at the boundary of the Brillouin zone (the signature

Load-bearing premise

The proof relies essentially on the assumption that the off-diagonal block G12 of the matrix symbol is normal and commutes with the diagonal block G11, so that the square of the full matrix G0 is block-diagonal; if that commutation fails, the lower-bound estimate on the resolvent that drives the convergence rates no longer holds.

Editorial extensions

If this is right

For Dirac operators in dimensions one, two, and three, the corrected symmetric-difference discretization converges in norm at rate O(h); for the bilayer graphene Hamiltonian, the rate is O(h^2).
The uncorrected symmetric-difference discretization never converges in norm for this class, so any numerical scheme based on it must include a correction to obtain spectral approximation guarantees.
The convergence implies Hausdorff convergence of spectra and essential spectra, and lower semicontinuity of discrete eigenvalues for perturbed operators, with the same rate for isolated eigenvalues.
The nonlocal square-root-of-Laplacian discretization converges without a correction term whenever the symbol is even in each momentum variable, offering a doubling-free alternative for such symbols.

Reading between the lines

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

The proof suggests a general design principle: any discretization whose symbol squares to a block-diagonal matrix with the correct lower bound will avoid fermion doubling; testing this on other lattice geometries (triangular, honeycomb) would be a natural next step.
The commutativity condition is restrictive but is satisfied by all supersymmetric-type symbols; the methods could likely be adapted to non-commuting blocks using a matrix-valued lower bound, at the cost of a weaker convergence rate.
The correction term's explicit form is flexible — only its abstract growth properties matter — so the proof should cover physically motivated corrections (e.g., Wilson fermions) as long as they satisfy the same symbol estimates.
The strong resolvent convergence of the uncorrected operator suggests that some spectral information (e.g., absolutely continuous spectrum) still converges even without a correction; quantifying exactly what is lost in norm versus retained strongly could refine practical guidance for lattice simulations.

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: the center holds, the tail doesn't. Theorem 3.7 (norm-resolvent convergence for the corrected operator and for the square-root discretization) is a genuine extension of Cornean–Garde–Jensen to 2×2 block matrices whose off-diagonal blocks are normal and commute with the diagonal block. The proof route is coherent: the diagonalization of G0^2 is correct, the Taylor estimates in Prop. 3.4 check out, and the aliasing sum reduces to |j|≤1 as claimed. Prop. 3.9 is a clean proof that the uncorrected symmetric-difference scheme fails in norm, so the correction is genuinely necessary. The bilayer graphene example and the Barker-convergence equivalence (Prop. 4.10) are new, and the Barker section is a useful clarification of the Weidmann/QUE landscape.

The soft spot is Theorem 3.10. The theorem says the uncorrected H_h still converges strongly even when (17) is dropped, but the proof uses estimate (19), which Lemma 3.3 proves only under (17). This is not a cosmetic detail. Take m=d=1, G12=0, G11=−a(1+ξ^2)^{1/2} with γ=1, β=0, and choose a so that at hξ=1 the term G11,h+f_h nearly vanishes. Then G_h^+ has an eigenvalue near zero and ∥D∥=O(1), while (19) would give O(h). The proof of (36) is invalid as written. The final bound may still be true because of the extra factor (G0±i)^{-1}, but the argument does not establish it. Either (17) must be reinstated or a direct proof is needed.

Second, smaller: Theorem 3.7 defers part of the heavy lifting to [6] with an 'obvious' matrix adaptation. In this genre that is often acceptable, but the referee should ask for the full matrix-valued version of [6, Lemma 3.3] to be written out, since the rest of the proof is otherwise careful.

The scope limitation (Assumptions 1(ii)–(iii)) is real but not hidden; the authors position the result as a supersymmetric-type class. The convergence rate is equivalent to [6] by their own Remark 3.6, so the novelty is in generality rather than exponents.

Verdict: send it to a serious referee. The paper deserves referee time; with Theorem 3.10 repaired or trimmed, it should be publishable. I would cite Theorem 3.7 and the Barker equivalence.

Referee Report

2 major / 4 minor

Summary. The paper studies continuum limits of discretized matrix-valued Fourier multipliers with block symbols G0 = [[G11, G12], [G12*, -G11]]. Under assumptions that G11 is self-adjoint, G12 is normal and commutes with G11, an elliptic lower bound holds, and derivatives grow at most polynomially, it proves norm-resolvent convergence rates for two discretizations: one based on the square-root of the standard second difference (tilde G_h) and one based on the symmetric difference with a correction term (G_h^+). For the symmetric difference without correction it proves, in Proposition 3.9, that generalized norm resolvent convergence fails, while Theorem 3.10 claims generalized strong resolvent convergence even when the additional positivity assumption (17) is dropped. Section 4 embeds these results in the frameworks of generalized Weidmann convergence, QUE-convergence, and Barker convergence.

Significance. If the main results hold, this is a substantial extension of the work of Cornean–Garde–Jensen: it covers a wide class of block operator matrices including Dirac and bilayer graphene symbols, gives explicit rates, and clarifies when a correction term is necessary. The proof of Theorem 3.7, in particular Proposition 3.4 and the aliasing reduction to |j|≤1, is coherent and the rates agree with the earlier scalar and Dirac results. Proposition 3.9 is a clean demonstration that the correction term is genuinely needed for norm convergence. The connections to QUE- and Barker-type convergence are useful exposition. The main caveat is that the proof of Theorem 3.10 uses an estimate that is only proved under the positivity assumption (17), although the theorem is explicitly advertised as dropping that assumption; this is a load-bearing gap in the paper as written.

major comments (2)

[Theorem 3.10, proof of (36)] The proof obtains (40), the bound for D(ξ)=(G_h^+(ξ)±i)^{-1}, from estimate (19). But (19) is proved in Lemma 3.3 only under Assumption 2, which includes (17) for G_h and G_h^+. The theorem is announced as valid 'even when the assumption (17) is dropped', so the proof is circular. The failure of (19) without (17) is concrete: take d=m=1, G12=0, G11(ξ)=-a(1+ξ^2)^{1/2} with a≈1.093. Assumptions 1 hold with γ=1, β=0, but G11 is not pointwise non-negative. At hξ=1, G11,h≈-0.918/h and f_h≈0.919/h, so G_h^+ has an eigenvalue near zero and ∥D(ξ)∥=O(1), while (19) would give O(h). Thus (40) is false without (17), and the proof of (36) is invalid as written. The final inequality (36) may still be true because of the extra right factor (G0±i)^{-1}, but that requires a different argument.
[Theorem 3.10, proof of (37)] The proof of the strong resolvent bound (37) inherits the same problem. For |j|=1 it bounds S_h(ξ+2πj/h) by '(C+Ch^γ)Ch^γ', using periodicity of G_h^+ combined with (40). Since (40) is not available without (17), the estimate ∥S_h∥≤Ch^γ in the aliasing region is unsupported. Consequently the strong convergence statement (37) and its consequence (38) are not established by the argument given. The final density step is itself fine, but it depends on (37).

minor comments (4)

[Eq. (16) and surrounding text] The definition f_h(ξ) = ( h ∑_{j=1}^m 4/h^2 sin^2(hξ_j/2) )^γ is correct for the symbol of (-hΔ_h)^γ but initially looks surprising because of the extra factor h. A one-line explanation that h times 4/h^2 is 4/h and that the small-ξ asymptotics is h^γ|ξ|^{2γ} would help readability.
[Introduction] The sentence 'F_h^* f_h F_h = -i d_h' and the parenthetical remark about the square for the second variant are terse. It would be clearer to state explicitly that for f_h(ξ)=sin(hξ)/h the corresponding operator is -i times the symmetric difference, and for f_h(ξ)=2/h |sin(hξ/2)| its square is -Δ_h.
[Proposition 3.9] The lower bound in (34) is valid because A and D are invertible, but the reader must supply the easy identity ∥B∥=∥A^{-1}(ABD)D^{-1}∥≤∥A^{-1}∥∥ABD∥∥D^{-1}∥. A short comment would avoid confusion.
[Section 4.2] In the two examples after Remark 4.12, the notation (χ_n)_j = δ_{j-1}^n is ambiguous; writing χ_n = δ_{j-1,n} or defining it explicitly would be clearer.

Circularity Check

1 steps flagged · score 4.0 of 10

Main theorems are independent, but Theorem 3.10's proof re-imports assumption (17) via estimate (19), undermining its claim to hold without (17).

other [Theorem 3.10 proof (p. 12), using (19) from Lemma 3.3 (p. 7) under Assumptions 2 (p. 7)]
"On the other hand, H_h still converges to H_0 in the generalized strong resolvent sense, even when the assumption (17) is dropped. ... Similarly, from (19) it follows that ∥D(ξ)∥ ≤ C(1+|ξ|^{2γ})^{-1/2}. [Lemma 3.3:] For the second inequality, we derive from (17) and non-negativity of f_h(ξ) that ..."

Theorem 3.10 purports to be valid without (17), but its proof of (36) bounds D(ξ)=(G_h^+(ξ)±i)^{-1} by (19)/(40). Lemma 3.3 obtains (19) only 'from (17) and non-negativity of f_h(ξ)'. Hence the proof re-imports exactly the assumption the theorem claims to drop. This is not merely cosmetic: without (17) one can arrange G11(ξ)≤0 so that at |ξ|~1/h the eigenvalue of G_h^+ is O(h), making ∥D∥=O(1) while (40) would give O(h). The final bound may still be true because of the extra factor (G0±i)^{-1}, but the argument as written is circular/incomplete.

full rationale

The central derivation is not circular: Theorem 3.7 is derived from Assumptions 1 and 2 via Lemma 3.3 and Proposition 3.4, and the heavy lifting from [6,7] is by other authors; [14] is used only for an example of backward/forward discretizations. The only genuinely self-referential step is in Theorem 3.10, where the advertised 'even when (17) is dropped' strong-convergence result is proved using estimate (19), whose proof in Lemma 3.3 requires (17). This is a real gap in a secondary theorem, but it does not reduce the paper's principal norm-resolvent-convergence claim to its own inputs, so the overall score is moderate rather than high.

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

The central claims rest on the model assumptions on G0 (normal commuting blocks, ellipticity exponent γ) and on the biorthogonal generating functions φ0,ψ0 with support conditions (6) imported from [6]. No numbers are fitted and no new physical entities are introduced; the correction term f_h is a constructive device determined by γ.

assumptions (5)

domain assumption Existence of biorthogonal Riesz generating functions φ0, ψ0 with h-independent constants and Fourier support conditions (6): supp φ̂0,ψ̂0 ⊂ [-3π/2,3π/2]^m and |φ̂0|,|ψ̂0| ≥ C0 on [-π/2,π/2]^m.
Borrowed from [6, Section 2]; used in Section 2.4 to define J_h,K_h and in Theorem 3.7 to restrict the aliasing sum (30) to |j|≤1.
domain assumption Assumptions 1(i)-(v) on G11,G12: G11 self-adjoint, G12 normal, [G11,G12]=0, elliptic lower bound G11^2+G12G12* ≥ C|ξ|^{2γ}I, derivative growth |∇(G⋆)ij| ≤ C|ξ|^β.
Defines the class of symbols; (ii)+(iii) make G0^2 block-diagonal, leading to the resolvent bound (10); (iv)+(v) give the decay used throughout Props. 3.4 and 3.9.
domain assumption Assumption 2: (13) G0(ξ)=G0(|ξ1|,...,|ξm|) for the \tilde G_h scheme, and (17) G11(ξ) ≥ 0 for the G_h/G+_h scheme.
(13) makes the substitution 2/h sin(hξ/2) consistent; (17) is used in Lemma 3.3(19) to get uniform resolvent decay for the corrected symbol.
standard math Standard Fourier analysis: Fourier–Plancherel transform unitarity, and the C*-identity ∥(A±i)^{-1}∥^2 = ∥(A^2+1)^{-1}∥ for self-adjoint A.
Used throughout, e.g. eq. (12), Prop. 3.4, Prop. 3.9.
standard math The identification operators J_h,K_h satisfy K_hJ_h = I and ∥J_h∥,∥K_h∥ bounded uniformly in h.
Follows from the Riesz property in Section 2.4; used in Prop. 3.9 to transfer pointwise lower bounds to operator norms.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Continuum limit of discretized matrix-valued Fourier multipliers." pith.science (2026). https://pith.science/paper/DVBGPL4W

@misc{pith2026260721177,
  author       = {Pith},
  title        = {Pith review of: Continuum limit of discretized matrix-valued Fourier multipliers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/DVBGPL4W}},
  note         = {Machine review of arXiv:2607.21177}
}

read the original abstract

Building upon a recent result by H. Cornean, H. Garde, and A. Jensen concerning continuum limits of discrete Dirac operators, we extend the analysis to a wide class of block operator matrices. This class includes, among others, the bilayer graphene Hamiltonian. Our main goal is to find norm estimates for the difference between the resolvents of continuous operators and their discrete counterparts embedded in the continuum in a specific way. While some discretization schemes lead directly to convergence in the generalized norm resolvent sense as the mesh parameter tends to zero, others require the addition of a suitable correction term to ensure the convergence.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 2 linked inside Pith

[6]

Cornean, H

H. Cornean, H. Garde, A. Jensen. Norm resolvent convergence of discretized Fourier multi- pliers. Journal of Fourier Analysis and Applications 27(4), 2021

2021
[1]

Bambusi, T

D. Bambusi, T. Penati. Continuous approximation of breathers in one-and two-dimensional DNLS lattices. Nonlinearity 23(1), 2010

2010
[2]

Journal of Functional Analysis 186(1), 2001

L.Barker.Continuumquantumsystemsaslimitsofdiscretequantumsystems, I:statevectors. Journal of Functional Analysis 186(1), 2001

2001
[3]

L. Barker. Continuum quantum systems as limits of discrete quantum systems: II. State functions. Journal of Physics A: Mathematical and General 34(22), 2001

2001
[4]

L. Barker. Continuum quantum systems as limits of discrete quantum systems. III. Operators. Journal of Mathematical Physics 42(10), 2001

2001
[5]

Vichnevetsky, J

R. Vichnevetsky, J. B. Bowles.Fourier analysis of numerical approximations of hyperbolic equations.Society for Industrial and Applied Mathematics, Philadelphia, 1982

1982
[7]

Cornean, H

H. Cornean, H. Garde, A. Jensen. Discrete approximations to Dirac operators and norm resolvent convergence. Journal of Spectral Theory 12(4), 2023

2023
[8]

Cornean, H

H. Cornean, H. Garde, A. Jensen. Discrete approximations to Dirichlet and Neumann Lapla- cians on a half-space and norm resolvent convergence. Studia Mathematica 271, 2023

2023

Show all 27 references

[9]

Exner, S

P. Exner, S. Nakamura, Y. Tadano. Continuum limit of the lattice quantum graph Hamil- tonian. Letters in Mathematical Physics 112(83), 2022

2022
[10]

International Journal of En- gineering Science 224, 2026

L.Fusi, O.Křenek, V.Průša, C.Rodriguez, R.Tozzi, M.Vejvoda.Discreteversuscontinuous– Linear lattice models and their exact continuous counterparts. International Journal of En- gineering Science 224, 2026

2026
[11]

Isozaki, A

H. Isozaki, A. Jensen. Continuum limit for lattice Schrödinger operators. Reviews in Mathe- matical Physics, 34(02), 2022

2022
[12]

Mikami, S

K. Mikami, S. Nakamura, Y. Tadano. Continuum limit for Laplace and elliptic operators on lattices. Pure and Applied Analysis 6(3), 2024

2024
[13]

Miranda, D

P. Miranda, D. Parra. Continuum limit for a discrete Hodge–Dirac operator on square lattices. Letters in Mathematical Physics 113(45), 2023

2023
[14]

Karapetyan

R. Karapetyan. Continuum limit for discrete block operator matrices. Diploma Thesis (Czech Technical University in Prague).http://hdl.handle.net/10467/123429, 2025

2025
[15]

Nakamura

S. Nakamura. Remarks on discrete Dirac operators and their continuum limits. Journal Spec- tral Theory 14(1), 2024

2024
[16]

Nakamura, Y

S. Nakamura, Y. Tadano. On a continuum limit of discrete Schrödinger operators on square lattice. Journal of Spectral Theory 11(1), 2021

2021
[17]

H. B. Nielsen, M. Ninomiya. A no-go theorem for regularizing chiral fermions. Physics Letters B 105(2–3), 1981

1981
[18]

Post.Spectral analysis on graph-like spaces

O. Post.Spectral analysis on graph-like spaces. Lecture Notes in Mathematics, vol. 2039, Springer, Heidelberg, 2012
[19]

O. Post. Spectral convergence of quasi-one-dimensional spaces. Annales Henri Poincaré 7, 2006

2006
[20]

O. Post, S. Zimmer. Generalised norm resolvent convergence: comparison of different con- cepts. Journal of Spectral Theory 12, 2022

2022
[21]

O. Post, S. Zimmer.Distances between operators acting on different Hilbert spaces, preprint athttps://arxiv.org/abs/2412.13165, 2025

2025 arXiv
[22]

H. J. Rothe.Lattice gauge theories: An introduction. Fourth edition, World Scientific Lecture Notes in Physics, vol. 82, World Scientific, Singapore, 2012

2012
[23]

K. M. Schmidt, T. Umeda. Continuum limits for discrete Dirac operators on 2D square lattices. Analysis and Mathematical Physics 13(46), 2023

2023
[24]

K. M. Schmidt, T. Umeda. On the resolvent convergence of discrete Dirac operators on 3D cubic lattices. arXiv preprint athttps://arxiv.org/abs/2507.01443. CONTINUUM LIMIT OF DISCRETIZED MATRIX-V ALUED FOURIER MULTIPLIERS 23

arXiv
[25]

Thaller.The Dirac Equation

B. Thaller.The Dirac Equation. Springer-Verlag, 1992

1992
[26]

Weidmann.Lineare Operatoren in Hilberträumen

J. Weidmann.Lineare Operatoren in Hilberträumen. B. G. Teubner, Stuttgart, 1976

1976
[27]

Weidmann

J. Weidmann. Stetige Abhängigkeit der Eigenwerte und Eigenfunktionen elliptischer Differ- entialoperatoren vom Gebiet. Math. Scand. 54 (1), 1984. Department of Computer Science, F aculty of Electrical Engineering, Czech Tech- nical University in Prague, Karlovo náměstí 13, 120...

1984

Pith tools

Reviewed August 1, 2026 · model on record in the stance chip above.

[1] [6]

Cornean, H

H. Cornean, H. Garde, A. Jensen. Norm resolvent convergence of discretized Fourier multi- pliers. Journal of Fourier Analysis and Applications 27(4), 2021

2021

[2] [1]

Bambusi, T

D. Bambusi, T. Penati. Continuous approximation of breathers in one-and two-dimensional DNLS lattices. Nonlinearity 23(1), 2010

2010

[3] [2]

Journal of Functional Analysis 186(1), 2001

L.Barker.Continuumquantumsystemsaslimitsofdiscretequantumsystems, I:statevectors. Journal of Functional Analysis 186(1), 2001

2001

[4] [3]

L. Barker. Continuum quantum systems as limits of discrete quantum systems: II. State functions. Journal of Physics A: Mathematical and General 34(22), 2001

2001

[5] [4]

L. Barker. Continuum quantum systems as limits of discrete quantum systems. III. Operators. Journal of Mathematical Physics 42(10), 2001

2001

[6] [5]

Vichnevetsky, J

R. Vichnevetsky, J. B. Bowles.Fourier analysis of numerical approximations of hyperbolic equations.Society for Industrial and Applied Mathematics, Philadelphia, 1982

1982

[7] [7]

Cornean, H

H. Cornean, H. Garde, A. Jensen. Discrete approximations to Dirac operators and norm resolvent convergence. Journal of Spectral Theory 12(4), 2023

2023

[8] [8]

Cornean, H

H. Cornean, H. Garde, A. Jensen. Discrete approximations to Dirichlet and Neumann Lapla- cians on a half-space and norm resolvent convergence. Studia Mathematica 271, 2023

2023