Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Saudi Journal of Engineering and Technology

How the Magnus Expansion Fails in Fermionic Systems: Breakdown Signatures in the Hubbard Model

Laraib-Ul- Nissa, Waqar Yousaf, Taniyat Kanwal

Published Sep 14, 2026• 0 citations
Publisher

Abstract

The Magnus expansion is attractive for periodically driven quantum systems because every finite truncation exponentiates an anti-Hermitian generator and therefore preserves unitarity. Its practical failure in interacting fermionic systems is nevertheless subtle: a low-order approximation may remain unitary while becoming progressively less accurate as additional terms are retained. We establish order-resolved breakdown signatures for a half-filled, periodically driven one-dimensional Fermi-Hubbard chain. The one-period propagator is computed numerically exactly in the fixed-particle sector and compared with Magnus approximants through fourth order. Failure is defined operationally by a successive-order error ratio (rₙ=εₙ/εₙ₋₁≥ 1), not by an unsupported claim of divergence of the infinite series. Across a raw (9×29) interaction-frequency grid ((U/J=0…,8), (ω/J=2…,16)), at least one added order loses improvement at 36.40% of the 261 sampled points. The smallest sampled frequency above which all retained orders improve throughout the remaining high-frequency tail rises from (ω/J=2.5) at (U=0) to (11.5) at (U/J=8). At (U/J=4), the highest (r₄=1) crossing occurs at (ω/J=7.719), followed by a monotone sampled tail from (ω/J=8). High-frequency error exponents approach the expected omitted-order scaling. A norm-based convergence certificate is consistently more conservative than the observed finite-order boundary. These results show that Magnus breakdown in the Hubbard model is order dependent, interaction shifted, and detectable before unitarity is lost.

DOI
10.36348/sjet.2026.v11i09.004

Related papers

Benchmarking the Magnus Expansion for Interaction Quenches in the Fermi-Hubbard Model: Exact Diagonalization on Small Clusters

Laraib-Ul- Nissa, Muhammad Abdullah, Waqar Yousaf

We investigate the nonequilibrium relaxation dynamics of the one-dimensional (1D) Fermi-Hubbard model subjected to abrupt, global interaction quenches. Specifically, we benchmark the convergence properties, structural accuracy, and algorithmic breakdown of the Magnus expansion against numerically exact results obtained via full Exact Diagonalization (ED) on small, periodic lattice clusters. By tracking the real-time evolution of local observables, double occupancy (doublon density), and many-body state fidelity metrics, we map out the validity bounds of the low-order Magnus series across weak, moderate, and strong interaction regimes. Our findings demonstrate that while the Magnus expansion provides an exceptionally accurate description of short-time coherent dynamics, rapid phase matching, and initial prethermalization tendencies, its convergence is fundamentally bottlenecked at longer timescales. This breakdown is driven by the rapid growth of multi-particle entanglement, non-local operator spreading via nested commutators, and the emergence of severe state-space fragmentation inherent to dense, strongly interacting many-body spectra.

Saudi Journal of Engineering and Technology•Jun 11, 2026•0 citations

Single particle properties of the two-dimensional Hubbard model for real frequencies at weak coupling: Breakdown of the Dyson series for partial self-energy expansions

B. D. E. McNiven, G. Todd Andrews, J. P. F. LeBlanc

We generate the perturbative expansion of the single particle Green's function and related self-energy for a half-filled single-band Hubbard model on a square lattice. We invoke algorithmic Matsubara integration to evaluate single-particle quantities for real and Matsubara frequencies and verify results through comparison to existing data on the Matsubara axis. With low-order expansions at weak coupling we observe a number of outcomes expected at higher orders: the opening of a gap, pseudogap behavior, and Fermi-surface reconstruction. Based on low-order perturbations, we consider the phase diagram that arises from truncated expansions of the self-energy and Green's function and their relation via the Dyson equation. From Matsubara axis data, we observe insulating behavior in direct expansions of the Green's function, whereas the same order of truncation of the self-energy produces metallic behavior. This observation is supported by additional calculations for real frequencies. We attribute this difference to the order in which diagrams are implicitly summed in the Dyson series. By separating the reducible and irreducible contributions at each order we show that the reducible diagrams implicitly summed in the Dyson equation lead to incorrect physics in the half-filled Hubbard model. Our observations for this particular case lead us to question the utility of the Dyson equation for any problem that shows a disparity between reducible and irreducible contributions to the expansion of the Green's function.

arXiv:2105.05357•Physical review. B./Physical review. B•Sep 10, 2021•14 citations

Ω TRINITΛS ΞNGINΞ Ver.6.0.0: Constraint-Preserving Projected Dynamics Invariant Subspaces, Gauge Leakage, and Local Unitary Evolution

scheduler4861

Abstract We develop Ω TRINITAS ENGINE v6, a mathematical and computational framework for constraint-preserving projected unitary dynamics, with particular emphasis on invariant physical subspaces, gauge-constraint leakage, local unitary evolution, and reproducible Hamiltonian lattice gauge theory. The central problem is to distinguish approximation error in discrete or product-formula time evolution from genuine violation of a prescribed physical constraint. For an orthogonal projector P defining the physical subspace and a unitary evolution U, the framework quantifies constraint leakage through ||(I-P)UP|| and establishes the exact identity ||PUx||^2 = ||x||^2 - ||(I-P)Ux||^2 for physical initial states. In finite-dimensional Hilbert spaces, exact preservation of the physical subspace is characterized by invariance under U, equivalently by vanishing leakage. For products of local unitaries, leakage admits an additive bound in terms of the individual factors, while Hamiltonian-generated evolution is controlled through commutators between the projector and the local generators. In particular, when every local Hamiltonian term preserves the Gauss-law physical subspace, finite-depth product-formula dynamics can preserve the gauge constraint exactly even when the approximation to the target continuous-time evolution remains imperfect. We investigate these results computationally using an explicitly constructed gauge-enriched successor model based on the 1+1-dimensional U(1) lattice Schwinger model with staggered fermions, periodic boundary conditions, finite electric-flux truncation, and Gauss-law reduction. Independent constructions of the physical basis and Hamiltonian agree exactly in the audited finite systems, with zero matrix and spectral discrepancies and no detected Gauss-violating hopping transitions. A symmetric local product formula exhibits approximately third-order single-step error and second-order fixed-time convergence while maintaining numerical unitarity near machine precision. Sparse low-energy calculations at fixed physical volume reproduce the massless Schwinger rest-gap benchmark: an even-power continuum extrapolation gives 0.5642377786 in units of the gauge coupling, compared with the exact continuum value 1/sqrt(pi) = 0.5641895835. In the representative electric-flux cutoff test, variations are below the reported numerical resolution. Momentum-resolved T2 sectors further reproduce the expected relativistic dispersion with strong numerical agreement at the lowest momenta and a larger, explicitly retained deviation at higher momentum. The numerical analysis constitutes finite-regulator calibration and continuum extrapolation rather than certification of a controlled continuum limit. Systematic uncertainties associated with fit family, fit range, finite physical volume, regulator ordering, and higher-momentum behavior remain distinct from the proved constraint-preservation results. This separation prevents numerical agreement with a known continuum theory from being promoted to a stronger mathematical or physical claim. The work explicitly distinguishes mathematical proof, finite computational verification, numerical observation, calibration, and physical interpretation. Agreement with the Schwinger model is treated as a nontrivial calibration of the gauge-enriched successor construction, not as a derivation of gauge theory from canonical Ω TRINITAS dynamics. The manuscript does not establish endogenous U(1) gauge emergence from canonical Ω TRINITAS v9 and does not claim correspondence between Ω TRINITAS and the microscopic dynamics of the actual universe. Negative results, failed constructions, limitations, and unresolved questions are retained as part of the epistemic record. Ω TRINITAS ENGINE v6 therefore provides a reproducible reference framework for projected quantum dynamics, invariant subspaces, Gauss-law preservation, constraint leakage, Hamiltonian lattice gauge theory, local unitary circuits, Trotter and product-formula evolution, sparse exact diagonalization, electric-flux truncation, continuum extrapolation, and momentum-sector spectroscopy. The canonical reference implementation and machine-readable numerical results are designed to preserve a strict distinction between proved statements, computationally verified finite cases, numerical evidence, open problems, and unsupported physical correspondence, providing an auditable foundation for subsequent mathematical and computational investigations.

•Aug 17, 2026•0 citations

Trimers in the extended Hubbard model

R. R. Montenegro-Filho, D. R. B. Silva, D. Cogollo +1 more

The Lieb theorem is a cornerstone of quantum magnetism theory in condensed matter. In this work, we investigate the instability of the Lieb insulating ferrimagnetic phase in the extended Hubbard model on a trimer chain at half-filling, with one electron per site, under increasing the nearest-neighbor Coulomb coupling $V$. Our results show that despite a noticeable increase in doublon density with $V$, the ferrimagnetic insulating phase remains robust up to the phase separation (PS) line, which is observed at $V\ensuremath{\gtrsim}U/4$, where $U$ is the local Coulomb repulsion. Above the PS line, one of the coexisting phases is primarily populated by doublons on one of the two sublattices of the chain. This phase coexists with a metallic, unsaturated ferromagnetic phase for $U\ensuremath{\gtrsim}t$, and with a singlet phase for $U\ensuremath{\lesssim}t$, where $t$ is the intratrimer hopping amplitude. We estimate the PS and the crossover lines with the help of density matrix renormalization group calculations.

arXiv:2606.30176•Physical review. B./Physical review. B•Feb 10, 2025•2 citations

The Magnus Expansion for Non-Hermitian Hamiltonians

Yair Mulian

The Magnus expansion offers a powerful geometric method to express a time-ordered exponential operator solution, typical of time-dependent linear differential equations, as an ordinary operatorial exponential. This representation possesses advantageous theoretical and qualitative features, such as preserving the underlying Lie algebraic structure of the system while remaining an exact solution to the original differential equation. For any finite-dimensional Hermitian Hamiltonian governing a quantum system, the standard Magnus expansion guarantees a manifestly unitary representation at every finite order of truncation. However, this foundational property is no longer preserved if the underlying Hamiltonian is non-Hermitian. In this work, we derive a generalized, unitarized expansion framework that systematically restores and maintains the property of geometric unitarity for all bounded finite-dimensional non-Hermitian Hamiltonians.

arXiv:2505.09559•Modern mathematical physics•Jul 24, 2026•2 citations

A Variational Framework for Time-Dependent Quantum Systems with Applications to Floquet Hamiltonians

Ibsal Assi, Meenu Kumari, J. P. F. LeBlanc

We introduce a variational framework for approximating the time-evolution operator $\hat{U}(t)$ within a physically motivated operator manifold, reformulating quantum dynamics as a tractable problem in operator space using stationary action principle. For periodically driven systems, the resulting approximate evolution operator directly yields an effective Floquet Hamiltonian, offering a non-perturbative alternative to conventional expansion-based methods. The framework is systematically improvable by enlarging the operator pool and naturally incorporates symmetries and physical constraints. When the operator manifold is chosen from the terms of a truncated Magnus expansion, the variational procedure effectively resums the Magnus series within the restricted space, significantly enhancing accuracy. We benchmark the approach on the driven Rabi model, the driven Lipkin-Meshkov-Glick model, and the one-dimensional driven Ising chain, yielding effective Floquet Hamiltonians that are systematically more accurate than low-order Magnus expansions, particularly in regimes where the latter converge poorly, and illustrating applicability to systems with exponentially large Hilbert spaces. Although we focus here on Floquet systems, the formalism applies equally to generic time-dependent Hamiltonians, providing a versatile tool for non-equilibrium quantum dynamics.

quant-pharXiv:2607.29418•Aug 3, 2026•0 citations
The Magnus expansion is attractive for periodically driven quantum systems because every finite truncation exponentiates an anti-Hermitian generator and therefore preserves unitarity. Its practical failure in interacting fermionic systems is nevertheless subtle: a low-order approximation may remain unitary while becoming progressively less accurate as additional terms are retained. We establish order-resolved breakdown signatures for a half-filled, periodically driven one-dimensional Fermi-Hubbard chain. The one-period propagator is computed numerically exactly in the fixed-particle sector and compared with Magnus approximants through fourth order. Failure is defined operationally by a successive-order error ratio (rₙ=εₙ/εₙ₋₁≥ 1), not by an unsupported claim of divergence of the infinite series. Across a raw (9×29) interaction-frequency grid ((U/J=0…,8), (ω/J=2…,16)), at least one added order loses improvement at 36.40% of the 261 sampled points. The smallest sampled frequency above which all retained orders improve throughout the remaining high-frequency tail rises from (ω/J=2.5) at (U=0) to (11.5) at (U/J=8). At (U/J=4), the highest (r₄=1) crossing occurs at (ω/J=7.719), followed by a monotone sampled tail from (ω/J=8). High-frequency error exponents approach the expected omitted-order scaling. A norm-based convergence certificate is consistently more conservative than the observed finite-order boundary. These results show that Magnus breakdown in the Hubbard model is order dependent, interaction shifted, and detectable before unitarity is lost.
We investigate the nonequilibrium relaxation dynamics of the one-dimensional (1D) Fermi-Hubbard model subjected to abrupt, global interaction quenches. Specifically, we benchmark the convergence properties, structural accuracy, and algorithmic breakdown of the Magnus expansion against numerically exact results obtained via full Exact Diagonalization (ED) on small, periodic lattice clusters. By tracking the real-time evolution of local observables, double occupancy (doublon density), and many-body state fidelity metrics, we map out the validity bounds of the low-order Magnus series across weak, moderate, and strong interaction regimes. Our findings demonstrate that while the Magnus expansion provides an exceptionally accurate description of short-time coherent dynamics, rapid phase matching, and initial prethermalization tendencies, its convergence is fundamentally bottlenecked at longer timescales. This breakdown is driven by the rapid growth of multi-particle entanglement, non-local operator spreading via nested commutators, and the emergence of severe state-space fragmentation inherent to dense, strongly interacting many-body spectra.
We generate the perturbative expansion of the single particle Green's function and related self-energy for a half-filled single-band Hubbard model on a square lattice. We invoke algorithmic Matsubara integration to evaluate single-particle quantities for real and Matsubara frequencies and verify results through comparison to existing data on the Matsubara axis. With low-order expansions at weak coupling we observe a number of outcomes expected at higher orders: the opening of a gap, pseudogap behavior, and Fermi-surface reconstruction. Based on low-order perturbations, we consider the phase diagram that arises from truncated expansions of the self-energy and Green's function and their relation via the Dyson equation. From Matsubara axis data, we observe insulating behavior in direct expansions of the Green's function, whereas the same order of truncation of the self-energy produces metallic behavior. This observation is supported by additional calculations for real frequencies. We attribute this difference to the order in which diagrams are implicitly summed in the Dyson series. By separating the reducible and irreducible contributions at each order we show that the reducible diagrams implicitly summed in the Dyson equation lead to incorrect physics in the half-filled Hubbard model. Our observations for this particular case lead us to question the utility of the Dyson equation for any problem that shows a disparity between reducible and irreducible contributions to the expansion of the Green's function.
Abstract We develop Ω TRINITAS ENGINE v6, a mathematical and computational framework for constraint-preserving projected unitary dynamics, with particular emphasis on invariant physical subspaces, gauge-constraint leakage, local unitary evolution, and reproducible Hamiltonian lattice gauge theory. The central problem is to distinguish approximation error in discrete or product-formula time evolution from genuine violation of a prescribed physical constraint. For an orthogonal projector P defining the physical subspace and a unitary evolution U, the framework quantifies constraint leakage through ||(I-P)UP|| and establishes the exact identity ||PUx||^2 = ||x||^2 - ||(I-P)Ux||^2 for physical initial states. In finite-dimensional Hilbert spaces, exact preservation of the physical subspace is characterized by invariance under U, equivalently by vanishing leakage. For products of local unitaries, leakage admits an additive bound in terms of the individual factors, while Hamiltonian-generated evolution is controlled through commutators between the projector and the local generators. In particular, when every local Hamiltonian term preserves the Gauss-law physical subspace, finite-depth product-formula dynamics can preserve the gauge constraint exactly even when the approximation to the target continuous-time evolution remains imperfect. We investigate these results computationally using an explicitly constructed gauge-enriched successor model based on the 1+1-dimensional U(1) lattice Schwinger model with staggered fermions, periodic boundary conditions, finite electric-flux truncation, and Gauss-law reduction. Independent constructions of the physical basis and Hamiltonian agree exactly in the audited finite systems, with zero matrix and spectral discrepancies and no detected Gauss-violating hopping transitions. A symmetric local product formula exhibits approximately third-order single-step error and second-order fixed-time convergence while maintaining numerical unitarity near machine precision. Sparse low-energy calculations at fixed physical volume reproduce the massless Schwinger rest-gap benchmark: an even-power continuum extrapolation gives 0.5642377786 in units of the gauge coupling, compared with the exact continuum value 1/sqrt(pi) = 0.5641895835. In the representative electric-flux cutoff test, variations are below the reported numerical resolution. Momentum-resolved T2 sectors further reproduce the expected relativistic dispersion with strong numerical agreement at the lowest momenta and a larger, explicitly retained deviation at higher momentum. The numerical analysis constitutes finite-regulator calibration and continuum extrapolation rather than certification of a controlled continuum limit. Systematic uncertainties associated with fit family, fit range, finite physical volume, regulator ordering, and higher-momentum behavior remain distinct from the proved constraint-preservation results. This separation prevents numerical agreement with a known continuum theory from being promoted to a stronger mathematical or physical claim. The work explicitly distinguishes mathematical proof, finite computational verification, numerical observation, calibration, and physical interpretation. Agreement with the Schwinger model is treated as a nontrivial calibration of the gauge-enriched successor construction, not as a derivation of gauge theory from canonical Ω TRINITAS dynamics. The manuscript does not establish endogenous U(1) gauge emergence from canonical Ω TRINITAS v9 and does not claim correspondence between Ω TRINITAS and the microscopic dynamics of the actual universe. Negative results, failed constructions, limitations, and unresolved questions are retained as part of the epistemic record. Ω TRINITAS ENGINE v6 therefore provides a reproducible reference framework for projected quantum dynamics, invariant subspaces, Gauss-law preservation, constraint leakage, Hamiltonian lattice gauge theory, local unitary circuits, Trotter and product-formula evolution, sparse exact diagonalization, electric-flux truncation, continuum extrapolation, and momentum-sector spectroscopy. The canonical reference implementation and machine-readable numerical results are designed to preserve a strict distinction between proved statements, computationally verified finite cases, numerical evidence, open problems, and unsupported physical correspondence, providing an auditable foundation for subsequent mathematical and computational investigations.
The Lieb theorem is a cornerstone of quantum magnetism theory in condensed matter. In this work, we investigate the instability of the Lieb insulating ferrimagnetic phase in the extended Hubbard model on a trimer chain at half-filling, with one electron per site, under increasing the nearest-neighbor Coulomb coupling VVV. Our results show that despite a noticeable increase in doublon density with VVV, the ferrimagnetic insulating phase remains robust up to the phase separation (PS) line, which is observed at $V\ensuremath{\gtrsim}U/4$, where UUU is the local Coulomb repulsion. Above the PS line, one of the coexisting phases is primarily populated by doublons on one of the two sublattices of the chain. This phase coexists with a metallic, unsaturated ferromagnetic phase for $U\ensuremath{\gtrsim}t$, and with a singlet phase for $U\ensuremath{\lesssim}t$, where ttt is the intratrimer hopping amplitude. We estimate the PS and the crossover lines with the help of density matrix renormalization group calculations.
The Magnus expansion offers a powerful geometric method to express a time-ordered exponential operator solution, typical of time-dependent linear differential equations, as an ordinary operatorial exponential. This representation possesses advantageous theoretical and qualitative features, such as preserving the underlying Lie algebraic structure of the system while remaining an exact solution to the original differential equation. For any finite-dimensional Hermitian Hamiltonian governing a quantum system, the standard Magnus expansion guarantees a manifestly unitary representation at every finite order of truncation. However, this foundational property is no longer preserved if the underlying Hamiltonian is non-Hermitian. In this work, we derive a generalized, unitarized expansion framework that systematically restores and maintains the property of geometric unitarity for all bounded finite-dimensional non-Hermitian Hamiltonians.
We introduce a variational framework for approximating the time-evolution operator U^(t)\hat{U}(t)U^(t) within a physically motivated operator manifold, reformulating quantum dynamics as a tractable problem in operator space using stationary action principle. For periodically driven systems, the resulting approximate evolution operator directly yields an effective Floquet Hamiltonian, offering a non-perturbative alternative to conventional expansion-based methods. The framework is systematically improvable by enlarging the operator pool and naturally incorporates symmetries and physical constraints. When the operator manifold is chosen from the terms of a truncated Magnus expansion, the variational procedure effectively resums the Magnus series within the restricted space, significantly enhancing accuracy. We benchmark the approach on the driven Rabi model, the driven Lipkin-Meshkov-Glick model, and the one-dimensional driven Ising chain, yielding effective Floquet Hamiltonians that are systematically more accurate than low-order Magnus expansions, particularly in regimes where the latter converge poorly, and illustrating applicability to systems with exponentially large Hilbert spaces. Although we focus here on Floquet systems, the formalism applies equally to generic time-dependent Hamiltonians, providing a versatile tool for non-equilibrium quantum dynamics.