arXiv:2607.10454
Krylov Complexity for Time-Dependent Hamiltonians
Amin Faraji Astaneh, Parsa Kafashi
Abstract
We investigate Krylov spread complexity for states evolving under time-dependent Hamiltonians. For periodically driven systems, we formulate the problem within Floquet theory and show how the Magnus expansion provides a systematic approximation when the Floquet Hamiltonian is not available in closed form. We then extend this framework beyond periodic driving and demonstrate that, in addition to the globally truncated Magnus expansion, a piecewise Magnus expansion provides a reliable method when the global expansion loses convergence or accuracy. Our results provide practical tools for analyzing complexity growth in a broad class of time-dependent quantum systems.
AI Summary
The paper studies how to compute state Krylov (spread) complexity under explicitly time-dependent Hamiltonians using effective generators obtained from Floquet theory and Magnus expansions. It treats periodically driven two-level systems and a harmonic oscillator with time-dependent frequency. For periodic driving, the Krylov basis is constructed from a Floquet Hamiltonian, giving a stroboscopic complexity. For $H_F=\mathbf h\cdot\boldsymbol\sigma$ and an initial $\sigma_z$ eigenstate, $C_K(t)=(h_x^2+h_y^2)\sin^2(|\mathbf h|t)/|\mathbf h|^2$. Circular driving is solved exactly in a rotating frame; linear driving is treated near resonance by the rotating-wave approximation and at high frequency by a dressed-frame Magnus expansion. The latter produces a Bessel-renormalized longitudinal field and a second-order transverse correction involving Bessel and Struve functions. For the oscillator, symplectic propagation determines Bogoliubov coefficients. In the chosen even-Fock Krylov chain, $C_K(t)=|\nu(t)|^2/2$. A sudden quench gives an exact oscillatory expression; smooth quenches are computed numerically by composing local Magnus propagators, recovering sudden-quench and approximately adiabatic behavior. The paper reviews the sufficient convergence condition $\int_0^t\|H(s)\|ds<\pi$. Piecewise truncation at order $m$ has stated accumulated error $O(t\Delta^m)$, assuming controlled stability factors. Limitations include global Magnus convergence and long-time accuracy, accumulated local errors, quasienergy-branch choices, and micromotion outside stroboscopic times. Near zeros of the leading Bessel term, higher Magnus orders determine residual dynamics. Extensions beyond the examples and to operator complexity remain future directions.
Error found: The soft-quench examples do not directly compare global and piecewise Magnus approximations in a regime where the global approximation fails. The reported soft-quench calculations omit step sizes, quantitative numerical errors, and refinement tests, leaving their claimed numerical reliability insufficiently documented. The oscillator step-size prescription reuses a Hamiltonian operator norm that is infinite for the harmonic oscillator, rather than explicitly establishing convergence using the finite-dimensional phase-space generator.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 27 pages, 4 figures, Refs. updated