arxiv: 2604.00306 · v2 · submitted 2026-03-31 · 🧮 math-ph · math.AP· math.MP

Recognition: unknown

Localised Davies generators for unbounded operators

Jeffrey Galkowski , Maciej Zworski

Authors on Pith no claims yet

Pith reviewed 2026-05-08 02:19 UTC · model gemini-3-flash-preview

classification 🧮 math-ph math.APmath.MP MSC 35S0547L9081Q20 PACS 03.65.Sq05.30.-d

keywords LindbladianDavies generatorGibbs statepseudodifferential operatorssemiclassical analysisquantum thermalization

0 comments

The pith

Localized Davies generators can be extended to unbounded operators, enabling quantum Gibbs sampling for continuous physical systems.

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

This paper establishes that a recently developed method for creating quantum Gibbs samplers—originally designed for finite-dimensional systems—is valid for unbounded operators like those found in quantum mechanics on continuous spaces. By using time-localized functions to filter state evolution, the authors construct a Lindbladian (a mathematical description of open system evolution) that naturally settles into a Gibbs state. This extension is crucial because it allows researchers to study how continuous systems reach thermal equilibrium without needing to know every energy level of the system beforehand. It provides a bridge between the discrete world of quantum information and the continuous world of traditional physics.

Core claim

The authors prove that the construction of localized Davies generators remains mathematically rigorous when applied to semiclassical pseudodifferential operators. They demonstrate that the resulting Lindbladian operator drives the system toward a stationary Gibbs state with an error that decays rapidly as the localization window in the frequency domain is refined. This bridges a gap between discrete quantum algorithms and the continuous operators typically used to model physical particles and fields, ensuring that the 'quantum Gibbs sampler' concept works for systems with infinite-dimensional Hilbert spaces.

What carries the argument

Semiclassical pseudodifferential operators. These are mathematical tools that relate quantum observables to classical phase space, enabling the construction to handle infinite-dimensional systems by using a functional calculus that controls energy scales and localization.

If this is right

Quantum Gibbs samplers can now be rigorously defined for particle systems in potential wells, not just finite-dimensional qubits.
The error in the stationary state can be precisely controlled by the decay properties of the Fourier transform of the localization function.
It provides a mathematical foundation for simulating thermalization in semiclassical limits where classical and quantum dynamics meet.
The construction allows for the study of Lindblad evolution in the context of classical/quantum correspondence for open systems.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

This could facilitate the development of new algorithms for preparing thermal states on continuous-variable quantum computers.
The approach might be adaptable to non-equilibrium steady states if the localization windows are modified to represent multiple baths at different temperatures.
It suggests that the locality of thermalization in phase space is a robust feature that survives the transition from discrete to continuous systems.

Load-bearing premise

The result relies on the existence of a robust functional calculus for the operators and specific fast-decaying properties of the localization window's frequency profile.

What would settle it

Calculating the stationary state for a specific unbounded system, like a quantum harmonic oscillator, and finding that the error exceeds the predicted bound determined by the localization function's Fourier transform.

read the original abstract

A classical Davies generator provides a Lindbladian for which the Gibbs state is stationary. Its construction involves precise knowledge of the Bohr spectrum or equivalently state evolution for all times. Recently Chen, Kastoryano and Gilyen proposed a construction involving localisation in time and carried out it out in the case of finite dimensional Hilbert spaces. The resulting generators are called quantum Gibbs samplers as the corresponding Lindblad is expected to settle to the Gibbs state. In this note, we show that the construction also works for classes of unbounded operators, including pseudodifferential operators used in the study of classical/quantum correspondence in Lindblad evolution.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

A rigorous extension of localized Davies generators to the unbounded, semiclassical regime that bridges the gap between quantum information and microlocal analysis.

read the letter

The punchline here is that Galkowski and Zworski have addressed the 'finite-dimensional' constraint of the Chen, Kastoryano, and Gilyen (CKG) framework. They prove that localized Davies generators work for unbounded semiclassical pseudodifferential operators, which is the right setting for continuous-variable systems and classical-quantum correspondence.

What the paper does well is the technical translation. Moving from finite matrices to unbounded operators on L2 spaces usually introduces domain issues and convergence headaches. The authors use standard but sophisticated microlocal tools to show that the construction is robust. Specifically, Theorem 3.3 provides a clear error bound for the Gibbs state's stationarity, tied directly to the decay of the window function’s Fourier transform. This is not just a hand-waving generalization; it's a formal proof that the localization trick remains valid in the semiclassical limit.

I noticed the stress-test concern regarding the density of the Bohr spectrum. While it is true that a dense spectrum would make a discrete implementation of the Lindbladian (summing over jump operators) difficult, the paper primarily develops the continuous integral formulation. In the semiclassical context, we often expect a continuous density of states anyway. The authors aren't proposing a specific numerical discretization scheme for every potential; they are establishing the operator-theoretic foundation. The 'soft spot' is simply that it is a short technical note—it establishes the framework and stops there, without exploring the convergence rates of specific discretized approximations.

This paper is for anyone working on the mathematical physics of open systems who needs a solid foundation for Gibbs samplers in continuous space. The math is tight, the authors are experts in this specific calculus, and the result is a useful piece of the puzzle for semiclassical simulation.

I recommend accepting this. It’s a serious, technically sound contribution that fills a clear gap in the literature.

Referee Report

3 major / 2 minor

Summary. The manuscript extends the construction of localized Davies generators—previously developed for finite-dimensional systems—to the setting of unbounded operators, specifically focusing on semiclassical pseudodifferential operators in the S(m) class. The authors show that by localizing the interaction in time using a window function f, one can construct a Lindbladian whose steady state approximates the Gibbs state without needing full knowledge of the Hamiltonian's Bohr spectrum. The central result is Theorem 3.3, which provides a rigorous bound on the stationarity of the Gibbs state under the localized generator, where the error is controlled by the decay of the Fourier transform of f outside the frequency set Omega.

Significance. This work is significant as it provides a rigorous bridge between the 'Quantum Gibbs Sampler' literature in quantum information and the mathematical framework of microlocal analysis used in quantum chemistry and physics. Extending these generators to unbounded operators is non-trivial due to the technicalities of functional calculus and the potentially dense Bohr spectrum of such operators. The paper identifies the appropriate class of symbols and the decay properties required for the windowing function to maintain approximation guarantees. This is a load-bearing step for developing thermal state preparation algorithms for continuous-variable systems.

major comments (3)

[§2.2, Eq. (2.5)] The construction of the Lindbladian L utilizes a sum over a discrete set of frequencies Omega. In the case of unbounded operators on L^2(R^d), the Bohr spectrum {E_i - E_j} is often dense. The manuscript needs to specify the requirements on the density of Omega relative to the support of f-hat. If Omega is an approximation of a continuous spectrum, the paper should clarify if the sum in Eq. (2.5) converges in the strong operator topology or if a truncation is required, and how that truncation affects the stationarity bound.
[§3.1, Theorem 3.3] The bound on the stationarity error ||L*(rho_beta)|| assumes that the Gibbs state rho_beta is trace-class. While this is true for symbols in S(m) with sufficient growth (e.g., elliptic symbols), the manuscript should explicitly state the conditions on the order of the symbol m and the inverse temperature beta that ensure the partition function Z = Tr(exp(-beta H)) is well-defined and that the operator exp(-beta H) belongs to the trace class.
[§3.2, Proof of Theorem 3.3] The derivation relies on the approximation of the identity via the window function f. However, there is a potential 'discretization error' between the integral-form generator and the discrete sum in Eq. (2.5). The proof should explicitly account for the error introduced by replacing the continuous frequency integral with a sum over Omega, especially if the spacing in Omega is not infinitely fine.

minor comments (2)

[§2.1] The semiclassical parameter 'h' is used throughout, but its scaling in the final error bound of Theorem 3.3 is not explicitly shown. It would be helpful to clarify if the error is O(h^infinity) or if it depends on h in a more complex way.
[§4] Adding a concrete example, such as the one-dimensional harmonic oscillator, would illustrate how the frequency set Omega is selected and how the cardinality of Omega scales with the desired precision.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for their constructive feedback and for recognizing the significance of our work in bridging semiclassical analysis and quantum Gibbs sampling. The comments regarding the spectral density and the trace-class conditions for the Gibbs state are very well-taken and will significantly improve the technical rigor of the manuscript. We have addressed each point below and will incorporate these changes into the revised version.

read point-by-point responses

Referee: [§2.2, Eq. (2.5)] The construction of the Lindbladian L utilizes a sum over a discrete set of frequencies Omega. In the case of unbounded operators on L^2(R^d), the Bohr spectrum {E_i - E_j} is often dense. The manuscript needs to specify the requirements on the density of Omega relative to the support of f-hat.

Authors: We agree that the nature of the Bohr spectrum is critical. In the context of the semiclassical operators $Op_h(S(m))$ considered in this paper, we typically assume the symbol $m$ is such that the operator has a purely discrete spectrum (e.g., elliptic symbols with confining potentials). Under these conditions, $\Omega$ is indeed a discrete set. We will revise Section 2.2 to explicitly state that we assume a discrete spectrum. Furthermore, we will clarify that the sum in Eq. (2.5) converges in the strong operator topology due to the rapid decay of $\hat{f}$—which acts as a filter—and the fact that the jump operators $A(\omega)$ are defined via the spectral projection of the interaction. For operators with truly continuous spectra, the sum would be replaced by an integral, but that is beyond the scope of this note. revision: yes
Referee: [§3.1, Theorem 3.3] The bound on the stationarity error ||L*(rho_beta)|| assumes that the Gibbs state rho_beta is trace-class. While this is true for symbols in S(m) with sufficient growth (e.g., elliptic symbols), the manuscript should explicitly state the conditions on the order of the symbol m and the inverse temperature beta that ensure the partition function Z = Tr(exp(-beta H)) is well-defined.

Authors: The referee is correct that the trace-class condition is essential for the normalization of the Gibbs state. For pseudodifferential operators, this is guaranteed if the symbol $m(z)$ grows sufficiently fast as $|z| \to \infty$, ensuring that the eigenvalues $\lambda_n$ satisfy a Weyl law that makes $\sum e^{-\beta \lambda_n}$ convergent. We will update Theorem 3.3 to include the explicit requirement that $m$ must be an order function such that $1/m$ is in $L^1(\mathbb{R}^{2d})$, or equivalent growth conditions, to ensure that $\rho_\beta$ is a well-defined density operator in the trace class. revision: yes
Referee: [§3.2, Proof of Theorem 3.3] The derivation relies on the approximation of the identity via the window function f. However, there is a potential 'discretization error' between the integral-form generator and the discrete sum in Eq. (2.5). The proof should explicitly account for the error introduced by replacing the continuous frequency integral with a sum over Omega.

Authors: We appreciate this comment, as it highlights a potential point of confusion. In the case of a discrete Bohr spectrum, the 'integral-form' generator (which uses a time-averaging window $f$) and the 'discrete sum' in Eq. (2.5) are actually identical. The integral $\int f(t) e^{itH} A e^{-itH} dt$ reduces to the sum $\sum_{\omega \in \Omega} \hat{f}(\omega) A(\omega)$ by the spectral theorem. Thus, there is no 'discretization error' in the mathematical definition of the generator itself. The only error arises from the fact that $\hat{f}$ is not perfectly supported on $\Omega = \{0\}$, which is precisely what Theorem 3.3 quantifies. We will clarify this in the proof of Theorem 3.3 to ensure the reader understands the identity is exact for discrete spectra. revision: partial

Circularity Check

0 steps flagged

Self-contained mathematical extension of localized Lindbladians to unbounded operators.

full rationale

The paper provides a rigorous mathematical derivation extending the 'localized Davies generator' construction from finite-dimensional systems to unbounded pseudodifferential operators. The central claim—that this construction leads to an approximate Gibbs sampler—is supported by a direct proof (Theorem 3.3) which bounds the stationarity error using the decay properties of the localization window's Fourier transform. The derivation utilizes standard semiclassical analysis tools and functional calculus for S(m) class operators. While the authors cite their own previous work on semiclassical Lindblad evolution, these citations provide technical background and motivation rather than serving as the logical basis for the new proof. The result is an independent mathematical verification of a construction's validity in a more general operator-theoretic setting, with no identified circularities or 'predictions' that are forced by definition or fitting.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The paper operates within standard semiclassical analysis and quantum Markovian dynamics without introducing new physical entities.

axioms (1)

domain assumption Existence of Gibbs state for the operator class
The paper assumes the Hamiltonian H is such that exp(-beta H) is trace-class to define the Gibbs state.

pith-pipeline@v0.9.0 · 6166 in / 1350 out tokens · 11394 ms · 2026-05-08T02:19:24.999874+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

20 extracted references · 7 canonical work pages · 1 internal anchor

[1]

Becker, C

S. Becker, C. Rouz\'e and R. Salzmann, Gibbs Sampling in infinite dimensions, in preparation
[2]

Benoist, N

T. Benoist, N. Cuneo, V. Jak s i\'c, and C-A. Pillet, On entropy production of repeated quantum measurements III. Quantum detailed balance. arXiv:2511.00910

work page arXiv
[3]

Breuer and F

H.P. Breuer and F. Petruccione. The Theory of Open Quantum Systems, Oxford, 2002

2002
[4]

Chen, M.J

C-F. Chen, M.J. Kastoryano, and A. Gily\'en, An efficient and exact noncommutative quantum Gibbs sampler, arXiv:2311.09207

work page arXiv
[5]

Cipriani, F

F. Cipriani, F. Fagnola, and J.M. Lindsay, Spectral analysis and Feller [roperty for quantum Ornstein–Uhlenbeck semigroups, Comm Math Phys 210 , 85--105 (2000)

2000
[6]

Davies, Markovian master equations, Comm

E.B. Davies, Markovian master equations, Comm. Math. Phys., 39 (1974), 91--110

1974
[7]

Davies, Markovian master equations

E.B. Davies, Markovian master equations. I\!I, Mathematische Annalen, 219 (1976), 147--158

1976
[8]

Davies, Quantum dynamical semigroups and the neutron diffusion equation, Reports on Mathematical Physics 11 (1977), 169--188

E.B. Davies, Quantum dynamical semigroups and the neutron diffusion equation, Reports on Mathematical Physics 11 (1977), 169--188

1977
[9]

Hern\'andez, D

F. Hern\'andez, D. Ranard, C.J. Riedel, Classical correspondence beyond the Ehrenfest time for open quantum systems with general Lindbladians. Comm. Math. Phys. 406, 4 (2025)

2025
[10]

Dimassi and J

M. Dimassi and J. Sj\"ostrand, Spectral Asymptotics in the Semi-Classical Limit , Cambridge U Press, 1999

1999
[11]

Z. Ding, B. Li, and L. Lin, Efficient Quantum Gibbs Samplers with Kubo–Martin–Schwinger Detailed Balance Condition. Commun. Math. Phys. 406 (2025), https://doi.org/10.1007/s00220-025-05235-3

work page doi:10.1007/s00220-025-05235-3 2025
[12]

Dyatlov and M

S. Dyatlov and M. Zworski, Mathematical Theory of Scattering Resonances,\/ AMS 2019, http://math.mit.edu/ dyatlov/res/

2019
[13]

Galkowski and M

J. Galkowski and M. Zworski, Classical--Quantum correspondence in Lindblad evolution, (with an appendix by Z. Huang and M. Zworski), J. Math. Phys., 66 (2025), Paper No. 091503

2025
[14]

Kato, Perturbation Theory for Linear Operators, Corrected second edition, Springer, 1980

T. Kato, Perturbation Theory for Linear Operators, Corrected second edition, Springer, 1980

1980
[15]

Li, Long time quantum-classical correspondence for open systems in trace norm, Ann

Z. Li, Long time quantum-classical correspondence for open systems in trace norm, Ann. Henri Poincar\'e (2025). https://doi.org/10.1007/s00023-025-01613-8

work page doi:10.1007/s00023-025-01613-8 2025
[16]

Ramkumar and M

A. Ramkumar and M. Soleimanifar, Mixing time of quantum Gibbs sampling for random sparse Hamiltonians, arXiv:2411.04454

work page arXiv
[17]

Rapid Mixing of Quantum Gibbs Samplers for Weakly-Interacting Quantum Systems

S. S mıd, R. Meister, M. Berta, and R. Bondesan, Rapid mixing of quantum Gibbs samplers for weakly-interacting quantum systems, arXiv:2510.04954

work page internal anchor Pith review Pith/arXiv arXiv
[18]

Smith, Lindblad evolution with subelliptic diffusion, arXiv:2601.04489

H. Smith, Lindblad evolution with subelliptic diffusion, arXiv:2601.04489

work page arXiv
[19]

Temme, Lower bounds to the spectral gap of Davies generators

K. Temme, Lower bounds to the spectral gap of Davies generators. J. Math. Phys. 54 (2013)

2013
[20]

Zworski, Semiclassical Analysis, Graduate Studies in Mathematics 138 , AMS, 2012

M. Zworski, Semiclassical Analysis, Graduate Studies in Mathematics 138 , AMS, 2012

2012